diff --git a/docs/api/openapi.yaml b/docs/api/openapi.yaml
index cdbce07fa..3153d95c6 100644
--- a/docs/api/openapi.yaml
+++ b/docs/api/openapi.yaml
@@ -24630,8 +24630,11 @@ paths:
       tags:
         - Anamnesis
       summary: Record an AUDIT-C (v1.43)
-      description: Scores the three items on the server. A replay with the same `externalId` answers 200 with the stored
-        result. 422 `anamnesis.auditC.invalid`; 429 `anamnesis.auditC.rateLimited` after 30 in an hour.
+      description: Scores the three items on the server. The entry is stored in the language of the wording the read route
+        offers this account (`auditC.instrument.locale`), and the cut-off follows that wording; `locale`, when sent,
+        must name that same wording. A replay with the same `externalId` answers 200 with the stored result. 422
+        `anamnesis.auditC.invalid` (also for another wording than the one offered, and for a `takenAt` more than five
+        minutes ahead of the server clock); 429 `anamnesis.auditC.rateLimited` after 30 in an hour.
       parameters:
         - *a6
       requestBody:
@@ -38945,6 +38948,9 @@ components:
               id:
                 type: string
                 minLength: 1
+              ref:
+                type: string
+                minLength: 1
               name:
                 type: string
                 minLength: 1
@@ -40524,6 +40530,9 @@ components:
               id:
                 type: string
                 minLength: 1
+              ref:
+                type: string
+                minLength: 1
               panel:
                 type:
                   - string
@@ -81038,11 +81047,12 @@ components:
         - standard
         - full
       description: "How much background AI work an account asks for, in ceiling order. `frugal`: status cards regenerate at
-        most every 24 h after a sync and every 6 h after a hand entry, a smaller snapshot, no background reasoning, a
-        short daily briefing (`recommendations` holds at most one entry, every field stays present), no automatic
-        document reading, no supplement label reading. `standard`: the operator's sync gap (default 6 h), 1 h after a
-        hand entry, background reasoning at the job default. `full`: 1 h after any ingest (the behaviour before v1.43).
-        The interactive Coach is the same at every level."
+        most every 24 h after a sync (or the operator's sync gap when that is longer, so never more often than
+        `standard`) and every 6 h after a hand entry, a smaller snapshot, no background reasoning, a short daily
+        briefing (`recommendations` holds at most one entry, every field stays present), no automatic document reading,
+        no supplement label reading. `standard`: the operator's sync gap (default 6 h), 1 h after a hand entry,
+        background reasoning at the job default. `full`: 1 h after any ingest (the behaviour before v1.43). The
+        interactive Coach is the same at every level."
     CoachReasoningState:
       type: object
       properties:
@@ -86447,6 +86457,7 @@ components:
               anyOf:
                 - $ref: "#/components/schemas/AuditCResult"
                 - type: "null"
+              description: The newest entry whose answers this host can open; an entry it cannot open is listed in `history` only.
             history:
               type: array
               items:
@@ -86470,6 +86481,7 @@ components:
               description: Newest first, at most 20.
             offerDue:
               type: boolean
+              description: No readable entry, or the newest one whose answers open (the one in `latest`) is 365 days old or older.
             instrument:
               type: object
               properties:
diff --git a/messages/de.json b/messages/de.json
index 421fcb35c..b89725bc1 100644
--- a/messages/de.json
+++ b/messages/de.json
@@ -8223,7 +8223,15 @@
           "busy": "Ein Probelauf oder eine Neuberechnung läuft noch. Versuche es erneut, sobald er fertig ist.",
           "dryRunRequired": "Starte zuerst einen Probelauf. Zum Anwenden braucht es einen, der in den letzten sieben Tagen abgeschlossen wurde.",
           "generic": "Die Anfrage ist nicht durchgegangen."
-        }
+        },
+        "rebuild": {
+          "queued": "Die Neuberechnung wartet auf den Hintergrunddienst.",
+          "running": "Die Pulstage werden neu berechnet.",
+          "done": "Pulstage am {at} für {accounts} Konten neu berechnet.",
+          "failed": "Die Neuberechnung ist abgebrochen. Das App-Log nennt die Ursache; starte sie erneut."
+        },
+        "rebuildAgain": "Erneut berechnen",
+        "rebuildStarted": "Neuberechnung gestartet."
       }
     },
     "carrier": "Netzbetreiber",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "Der Eintrag verschwindet aus deinem Verlauf. Das lässt sich hier nicht rückgängig machen.",
         "removedToast": "Eintrag entfernt",
         "removeError": "Der Eintrag konnte nicht entfernt werden.",
-        "historyEntry": "{date}: {score} von 12"
+        "historyEntry": "{date}: {score} von 12",
+        "legacyLabel": "Frühere Angabe: {value}",
+        "legacyHint": "Aus der kurzen Alkoholfrage vor diesem Fragebogen. Sobald ein Fragebogen-Ergebnis vorliegt, liest der Coach sie nicht mehr.",
+        "legacyRemove": "Frühere Angabe entfernen",
+        "legacyRemoveConfirmTitle": "Frühere Angabe entfernen?",
+        "legacyRemoveConfirmBody": "Die frühere Angabe verschwindet aus deinen Daten. Fragebogen-Ergebnisse bleiben.",
+        "legacyRemovedToast": "Frühere Angabe entfernt",
+        "legacyRemoveError": "Die frühere Angabe konnte nicht entfernt werden."
       },
       "takeOver": {
         "conditionsButton": "In Zustandstagebuch übernehmen",
diff --git a/messages/en.json b/messages/en.json
index b80e0a938..b20d52ba7 100644
--- a/messages/en.json
+++ b/messages/en.json
@@ -8223,7 +8223,15 @@
           "busy": "A dry run or recalculation is still running. Try again once it has finished.",
           "dryRunRequired": "Run a dry run first. Applying needs one that finished in the last seven days.",
           "generic": "The request did not go through."
-        }
+        },
+        "rebuild": {
+          "queued": "The recalculation is waiting for the background worker.",
+          "running": "The pulse days are being recalculated.",
+          "done": "Pulse days recalculated on {at} for {accounts} accounts.",
+          "failed": "The recalculation stopped. The app log names the cause; run it again."
+        },
+        "rebuildAgain": "Recalculate again",
+        "rebuildStarted": "Recalculation started."
       }
     },
     "carrier": "Carrier",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "The entry leaves your history. This cannot be undone here.",
         "removedToast": "Entry removed",
         "removeError": "The entry could not be removed.",
-        "historyEntry": "{date}: {score} of 12"
+        "historyEntry": "{date}: {score} of 12",
+        "legacyLabel": "Earlier answer: {value}",
+        "legacyHint": "From the short alcohol question asked before this questionnaire. Once a questionnaire result is on file, the Coach no longer reads it.",
+        "legacyRemove": "Remove earlier answer",
+        "legacyRemoveConfirmTitle": "Remove the earlier answer?",
+        "legacyRemoveConfirmBody": "The earlier answer leaves your record. Questionnaire results stay.",
+        "legacyRemovedToast": "Earlier answer removed",
+        "legacyRemoveError": "The earlier answer could not be removed."
       },
       "takeOver": {
         "conditionsButton": "Add to condition journal",
diff --git a/messages/es.json b/messages/es.json
index bb81a534a..3dde73720 100644
--- a/messages/es.json
+++ b/messages/es.json
@@ -8223,7 +8223,15 @@
           "busy": "Todavía hay una prueba o un recálculo en marcha. Inténtalo de nuevo cuando termine.",
           "dryRunRequired": "Ejecuta primero una prueba. Para aplicar hace falta una terminada en los últimos siete días.",
           "generic": "La solicitud no se completó."
-        }
+        },
+        "rebuild": {
+          "queued": "El recálculo espera al proceso en segundo plano.",
+          "running": "Se están recalculando los días de pulso.",
+          "done": "Días de pulso recalculados el {at} para {accounts} cuentas.",
+          "failed": "El recálculo se detuvo. El registro de la app indica la causa; vuelve a ejecutarlo."
+        },
+        "rebuildAgain": "Recalcular de nuevo",
+        "rebuildStarted": "Recálculo iniciado."
       }
     },
     "carrier": "Operador",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "La entrada sale de tu historial. Aquí no se puede deshacer.",
         "removedToast": "Entrada eliminada",
         "removeError": "No se pudo eliminar la entrada.",
-        "historyEntry": "{date}: {score} de 12"
+        "historyEntry": "{date}: {score} de 12",
+        "legacyLabel": "Respuesta anterior: {value}",
+        "legacyHint": "De la pregunta breve sobre alcohol anterior a este cuestionario. En cuanto hay un resultado del cuestionario, el Coach deja de leerla.",
+        "legacyRemove": "Eliminar respuesta anterior",
+        "legacyRemoveConfirmTitle": "¿Eliminar la respuesta anterior?",
+        "legacyRemoveConfirmBody": "La respuesta anterior sale de tu registro. Los resultados del cuestionario se mantienen.",
+        "legacyRemovedToast": "Respuesta anterior eliminada",
+        "legacyRemoveError": "No se pudo eliminar la respuesta anterior."
       },
       "takeOver": {
         "conditionsButton": "Pasar al diario de afecciones",
diff --git a/messages/fr.json b/messages/fr.json
index 96777e4a8..f53f6dc32 100644
--- a/messages/fr.json
+++ b/messages/fr.json
@@ -8223,7 +8223,15 @@
           "busy": "Une simulation ou un recalcul est encore en cours. Réessayez une fois terminé.",
           "dryRunRequired": "Lancez d'abord une simulation. L'application en exige une terminée au cours des sept derniers jours.",
           "generic": "La requête n'a pas abouti."
-        }
+        },
+        "rebuild": {
+          "queued": "Le recalcul attend le processus en arrière-plan.",
+          "running": "Les jours de pouls sont en cours de recalcul.",
+          "done": "Jours de pouls recalculés le {at} pour {accounts} comptes.",
+          "failed": "Le recalcul s’est arrêté. Le journal de l’app en indique la cause ; relancez-le."
+        },
+        "rebuildAgain": "Recalculer à nouveau",
+        "rebuildStarted": "Recalcul lancé."
       }
     },
     "carrier": "Opérateur",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "L'entrée quitte votre historique. Impossible d'annuler ici.",
         "removedToast": "Entrée supprimée",
         "removeError": "Impossible de supprimer l'entrée.",
-        "historyEntry": "{date} : {score} sur 12"
+        "historyEntry": "{date} : {score} sur 12",
+        "legacyLabel": "Réponse précédente : {value}",
+        "legacyHint": "Issue de la courte question sur l’alcool posée avant ce questionnaire. Dès qu’un résultat du questionnaire existe, le Coach ne la lit plus.",
+        "legacyRemove": "Supprimer la réponse précédente",
+        "legacyRemoveConfirmTitle": "Supprimer la réponse précédente ?",
+        "legacyRemoveConfirmBody": "La réponse précédente quitte votre dossier. Les résultats du questionnaire restent.",
+        "legacyRemovedToast": "Réponse précédente supprimée",
+        "legacyRemoveError": "La réponse précédente n’a pas pu être supprimée."
       },
       "takeOver": {
         "conditionsButton": "Reprendre dans le journal des affections",
diff --git a/messages/it.json b/messages/it.json
index 4acd04e2f..9ec0dc003 100644
--- a/messages/it.json
+++ b/messages/it.json
@@ -8223,7 +8223,15 @@
           "busy": "Una prova o un ricalcolo è ancora in corso. Riprova quando è terminato.",
           "dryRunRequired": "Avvia prima una prova. Per applicare ne serve una terminata negli ultimi sette giorni.",
           "generic": "La richiesta non è andata a buon fine."
-        }
+        },
+        "rebuild": {
+          "queued": "Il ricalcolo attende il processo in background.",
+          "running": "I giorni del polso sono in fase di ricalcolo.",
+          "done": "Giorni del polso ricalcolati il {at} per {accounts} account.",
+          "failed": "Il ricalcolo si è interrotto. Il log dell’app indica la causa; avvialo di nuovo."
+        },
+        "rebuildAgain": "Ricalcola di nuovo",
+        "rebuildStarted": "Ricalcolo avviato."
       }
     },
     "carrier": "Operatore",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "La voce esce dalla tua cronologia. Qui non si può annullare.",
         "removedToast": "Voce rimossa",
         "removeError": "Impossibile rimuovere la voce.",
-        "historyEntry": "{date}: {score} su 12"
+        "historyEntry": "{date}: {score} su 12",
+        "legacyLabel": "Risposta precedente: {value}",
+        "legacyHint": "Dalla breve domanda sull’alcol posta prima di questo questionario. Quando c’è un risultato del questionario, il Coach non la legge più.",
+        "legacyRemove": "Rimuovi risposta precedente",
+        "legacyRemoveConfirmTitle": "Rimuovere la risposta precedente?",
+        "legacyRemoveConfirmBody": "La risposta precedente esce dai tuoi dati. I risultati del questionario restano.",
+        "legacyRemovedToast": "Risposta precedente rimossa",
+        "legacyRemoveError": "Non è stato possibile rimuovere la risposta precedente."
       },
       "takeOver": {
         "conditionsButton": "Riporta nel diario delle condizioni",
diff --git a/messages/ko.json b/messages/ko.json
index 1bfd80398..d35f2619b 100644
--- a/messages/ko.json
+++ b/messages/ko.json
@@ -8198,7 +8198,15 @@
           "busy": "시험 실행이나 재계산이 아직 진행 중입니다. 끝난 뒤 다시 시도하세요.",
           "dryRunRequired": "먼저 시험 실행을 하세요. 적용하려면 최근 7일 안에 끝난 시험 실행이 필요합니다.",
           "generic": "요청이 처리되지 않았습니다."
-        }
+        },
+        "rebuild": {
+          "queued": "재계산이 백그라운드 작업을 기다리고 있습니다.",
+          "running": "맥박 일별 값을 재계산하고 있습니다.",
+          "done": "{at}에 계정 {accounts}개의 맥박 일별 값을 재계산했습니다.",
+          "failed": "재계산이 중단되었습니다. 앱 로그에 원인이 있습니다. 다시 실행하세요."
+        },
+        "rebuildAgain": "다시 계산",
+        "rebuildStarted": "재계산을 시작했습니다."
       }
     },
     "carrier": "통신사",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "기록이 내역에서 사라져요. 여기서는 되돌릴 수 없어요.",
         "removedToast": "기록을 삭제했어요",
         "removeError": "기록을 삭제하지 못했어요.",
-        "historyEntry": "{date}: 12점 중 {score}점"
+        "historyEntry": "{date}: 12점 중 {score}점",
+        "legacyLabel": "이전 답변: {value}",
+        "legacyHint": "이 설문 전에 묻던 짧은 음주 질문의 답변입니다. 설문 결과가 있으면 코치는 더 이상 이 답변을 읽지 않습니다.",
+        "legacyRemove": "이전 답변 삭제",
+        "legacyRemoveConfirmTitle": "이전 답변을 삭제할까요?",
+        "legacyRemoveConfirmBody": "이전 답변이 기록에서 삭제됩니다. 설문 결과는 그대로 남습니다.",
+        "legacyRemovedToast": "이전 답변을 삭제했습니다",
+        "legacyRemoveError": "이전 답변을 삭제하지 못했습니다."
       },
       "takeOver": {
         "conditionsButton": "상태 일지로 옮기기",
diff --git a/messages/pl.json b/messages/pl.json
index 8f35637a7..3284806be 100644
--- a/messages/pl.json
+++ b/messages/pl.json
@@ -8223,7 +8223,15 @@
           "busy": "Próba lub przeliczanie wciąż trwa. Spróbuj ponownie, gdy się zakończy.",
           "dryRunRequired": "Najpierw uruchom próbę. Do zastosowania potrzebna jest próba zakończona w ciągu ostatnich siedmiu dni.",
           "generic": "Żądanie nie zostało zrealizowane."
-        }
+        },
+        "rebuild": {
+          "queued": "Ponowne przeliczenie czeka na proces w tle.",
+          "running": "Dni pulsu są przeliczane.",
+          "done": "Dni pulsu przeliczone {at} dla kont: {accounts}.",
+          "failed": "Przeliczanie zostało przerwane. Przyczynę podaje log aplikacji; uruchom je ponownie."
+        },
+        "rebuildAgain": "Przelicz ponownie",
+        "rebuildStarted": "Przeliczanie rozpoczęte."
       }
     },
     "carrier": "Operator",
@@ -11279,7 +11287,14 @@
         "removeConfirmBody": "Wpis zniknie z historii. Tutaj nie da się tego cofnąć.",
         "removedToast": "Wpis usunięty",
         "removeError": "Nie udało się usunąć wpisu.",
-        "historyEntry": "{date}: {score} z 12"
+        "historyEntry": "{date}: {score} z 12",
+        "legacyLabel": "Wcześniejsza odpowiedź: {value}",
+        "legacyHint": "Z krótkiego pytania o alkohol sprzed tego kwestionariusza. Gdy jest wynik kwestionariusza, Coach przestaje ją czytać.",
+        "legacyRemove": "Usuń wcześniejszą odpowiedź",
+        "legacyRemoveConfirmTitle": "Usunąć wcześniejszą odpowiedź?",
+        "legacyRemoveConfirmBody": "Wcześniejsza odpowiedź zniknie z Twoich danych. Wyniki kwestionariusza zostają.",
+        "legacyRemovedToast": "Wcześniejsza odpowiedź usunięta",
+        "legacyRemoveError": "Nie udało się usunąć wcześniejszej odpowiedzi."
       },
       "takeOver": {
         "conditionsButton": "Przenieś do dziennika stanów",
diff --git a/src/app/api/admin/maintenance/pulse-day-mean/__tests__/route.test.ts b/src/app/api/admin/maintenance/pulse-day-mean/__tests__/route.test.ts
index 3f05cb0a1..26c89266e 100644
--- a/src/app/api/admin/maintenance/pulse-day-mean/__tests__/route.test.ts
+++ b/src/app/api/admin/maintenance/pulse-day-mean/__tests__/route.test.ts
@@ -66,10 +66,14 @@ const post = (body: unknown) =>
 const NOW = new Date("2026-10-16T12:00:00Z");
 const send = vi.fn();
 
-function settings(dryRun: unknown, method = "hour_mean_v1") {
+function settings(
+  dryRun: unknown,
+  method = "hour_mean_v1",
+  appliedAt: Date | null = null,
+) {
   vi.mocked(prisma.appSettings.findUnique).mockResolvedValue({
     pulseDayMeanMethod: method,
-    pulseDayMeanAppliedAt: null,
+    pulseDayMeanAppliedAt: appliedAt,
     pulseDayMeanDryRunJson: dryRun,
   } as never);
 }
@@ -136,6 +140,7 @@ describe("GET", () => {
       method: "hour_mean_v1",
       appliedAt: null,
       lastDryRun: doneAt("2026-10-15T08:00:00.000Z"),
+      lastRebuild: null,
     });
   });
 
@@ -199,7 +204,9 @@ describe("POST dry_run", () => {
     const res = await post({ action: "dry_run" });
     expect(res.status).toBe(409);
     expect(vi.mocked(prisma.appSettings.upsert).mock.lastCall?.[0]).toEqual(
-      expect.objectContaining({ update: { pulseDayMeanDryRunJson: previous } }),
+      expect.objectContaining({
+        update: { pulseDayMeanDryRunJson: { ...previous, rebuild: null } },
+      }),
     );
   });
 });
@@ -264,10 +271,71 @@ describe("POST apply", () => {
   });
 });
 
+describe("POST rebuild", () => {
+  const failedRebuild = {
+    status: "failed",
+    startedAt: "2026-10-15T09:00:00.000Z",
+    finishedAt: "2026-10-15T09:05:00.000Z",
+    accounts: null,
+    rows: null,
+  };
+
+  it("is refused before any apply", async () => {
+    settings(doneAt("2026-10-15T08:00:00.000Z"));
+    const res = await post({ action: "rebuild" });
+    expect(res.status).toBe(409);
+    expect((await res.json()).meta.errorCode).toBe(
+      "maintenance.pulseDayMean.dryRunRequired",
+    );
+    expect(send).not.toHaveBeenCalled();
+  });
+
+  it("runs the recalculation again after a stopped one, without a new dry run", async () => {
+    // The dry run is three weeks old: apply would refuse it, a rebuild does not.
+    settings(
+      { ...doneAt("2026-09-20T08:00:00.000Z"), rebuild: failedRebuild },
+      "weighted_hour_mean_v2",
+      new Date("2026-09-20T09:00:00.000Z"),
+    );
+    const res = await post({ action: "rebuild" });
+    expect(res.status).toBe(202);
+    expect(send).toHaveBeenCalledWith(
+      "pulse-day-mean",
+      { action: "rebuild" },
+      expect.anything(),
+    );
+    expect(
+      vi.mocked(prisma.appSettings.upsert).mock.lastCall?.[0],
+    ).toMatchObject({
+      update: {
+        pulseDayMeanDryRunJson: { rebuild: { status: "queued" } },
+      },
+    });
+    expect(auditLog).toHaveBeenCalledWith(
+      "admin.pulseDayMean.rebuild",
+      expect.anything(),
+    );
+  });
+
+  it("shows a rebuild left queued with no job behind it as failed", async () => {
+    settings(
+      {
+        ...doneAt("2026-10-15T08:00:00.000Z"),
+        rebuild: { ...failedRebuild, status: "running", finishedAt: null },
+      },
+      "weighted_hour_mean_v2",
+      new Date("2026-10-15T09:00:00.000Z"),
+    );
+    const body = await (await get()).json();
+    expect(body.data.lastRebuild.status).toBe("failed");
+    expect(body.data.lastDryRun.status).toBe("done");
+  });
+});
+
 describe("POST body", () => {
   it("is strict", async () => {
     expect((await post({ action: "apply", force: true })).status).toBe(422);
-    expect((await post({ action: "rebuild" })).status).toBe(422);
+    expect((await post({ action: "purge" })).status).toBe(422);
     expect(send).not.toHaveBeenCalled();
   });
 
diff --git a/src/app/api/admin/maintenance/pulse-day-mean/route.ts b/src/app/api/admin/maintenance/pulse-day-mean/route.ts
index c94e34f50..12e0ebf5d 100644
--- a/src/app/api/admin/maintenance/pulse-day-mean/route.ts
+++ b/src/app/api/admin/maintenance/pulse-day-mean/route.ts
@@ -4,7 +4,7 @@
  *
  * GET answers the active method, when it was applied and the last dry run
  * (status and result: per-account ordinal labels only, never an id). POST
- * `{ action: "dry_run" | "apply" }` answers 202 `{ enqueued }`: a dry run
+ * `{ action: "dry_run" | "apply" | "rebuild" }` answers 202 `{ enqueued }`: a dry run
  * computes both methods for every account and day and writes nothing but its
  * result; `apply` needs a successful dry run from the last seven days (409
  * `maintenance.pulseDayMean.dryRunRequired`), switches the method and
@@ -31,6 +31,7 @@ import {
   PULSE_DAY_MEAN_SEND_OPTIONS,
   readPulseDayMeanStatus,
   writeDryRun,
+  writeRebuild,
   type PulseDayMeanPayload,
 } from "@/lib/jobs/pulse-day-mean";
 import { annotate } from "@/lib/logging/context";
@@ -41,6 +42,7 @@ import {
 import {
   hasFreshDryRun,
   pulseDayMeanActionSchema,
+  type PulseDayMeanStatus,
 } from "@/lib/validations/pulse-day-mean";
 
 export const dynamic = "force-dynamic";
@@ -54,6 +56,33 @@ function busy() {
   });
 }
 
+/**
+ * Queue the rebuild and record it as queued, so the card shows it from the
+ * moment it is asked for and a rebuild that never starts reads as failed.
+ * When another run holds the singleton, the previous rebuild state is put
+ * back and the answer is false.
+ */
+async function sendRebuild(
+  boss: NonNullable<ReturnType<typeof getGlobalBoss>>,
+  previous: PulseDayMeanStatus["lastRebuild"],
+): Promise<boolean> {
+  await writeRebuild({
+    status: "queued",
+    startedAt: new Date().toISOString(),
+    finishedAt: null,
+    accounts: null,
+    rows: null,
+  });
+  const payload: PulseDayMeanPayload = { action: "rebuild" };
+  const jobId = await boss.send(
+    PULSE_DAY_MEAN_QUEUE,
+    payload,
+    PULSE_DAY_MEAN_SEND_OPTIONS,
+  );
+  if (jobId === null) await writeRebuild(previous);
+  return jobId !== null;
+}
+
 export const GET = apiHandler(async () => {
   await requireAdmin();
   const status = await readPulseDayMeanStatus();
@@ -110,6 +139,28 @@ export const POST = apiHandler(async (request: NextRequest) => {
     return apiSuccess({ enqueued: true }, 202);
   }
 
+  if (action === "rebuild") {
+    // The recalculation again, after an apply whose rebuild stopped. The
+    // method is already switched; the dry run that justified it stands.
+    const status = await readPulseDayMeanStatus();
+    if (status.appliedAt === null) {
+      return apiError(
+        "Nothing applied yet; run a dry run and apply first",
+        409,
+        { errorCode: "maintenance.pulseDayMean.dryRunRequired" },
+      );
+    }
+    const enqueued = await sendRebuild(boss, status.lastRebuild);
+    if (!enqueued) return busy();
+    await auditLog("admin.pulseDayMean.rebuild", {
+      userId: user.id,
+      ipAddress: getClientIp(request),
+      details: { method: status.method },
+    });
+    annotate({ meta: { pulse_day_mean_enqueued: true } });
+    return apiSuccess({ enqueued: true }, 202);
+  }
+
   if (!hasFreshDryRun(await readPulseDayMeanStatus())) {
     return apiError(
       "Run a dry run first; apply needs one finished in the last seven days",
@@ -135,13 +186,10 @@ export const POST = apiHandler(async (request: NextRequest) => {
     },
   });
   setActivePulseDayMeanMethod(APPLIED_METHOD);
-  const payload: PulseDayMeanPayload = { action: "rebuild" };
-  const jobId = await boss.send(
-    PULSE_DAY_MEAN_QUEUE,
-    payload,
-    PULSE_DAY_MEAN_SEND_OPTIONS,
+  const enqueued = await sendRebuild(
+    boss,
+    (await readPulseDayMeanStatus()).lastRebuild,
   );
-  const enqueued = jobId !== null;
 
   await auditLog("admin.pulseDayMean.apply", {
     userId: user.id,
diff --git a/src/app/api/anamnesis/audit-c/route.ts b/src/app/api/anamnesis/audit-c/route.ts
index fa979b285..cae7171d6 100644
--- a/src/app/api/anamnesis/audit-c/route.ts
+++ b/src/app/api/anamnesis/audit-c/route.ts
@@ -26,10 +26,14 @@ import {
 import { auditLog } from "@/lib/auth/audit";
 import { withIdempotency } from "@/lib/idempotency";
 import { annotate } from "@/lib/logging/context";
+import { auditCInstrument } from "@/lib/mental-health/audit-c";
 import { createAuditC } from "@/lib/profile/life-circumstances";
 import { checkRateLimit } from "@/lib/rate-limit";
 import { auditCRequestSchema } from "@/lib/validations/life-circumstances";
 
+/** How far ahead of the server clock a `takenAt` may lie. */
+const AUDIT_C_CLOCK_SKEW_MS = 5 * 60 * 1000;
+
 export const POST = apiHandler(withIdempotency<[NextRequest]>(postAuditC));
 
 async function postAuditC(request: NextRequest): Promise<Response> {
@@ -54,11 +58,36 @@ async function postAuditC(request: NextRequest): Promise<Response> {
   }
 
   const input = parsed.data;
+
+  // The cut-off follows the language of the wording that was answered, and
+  // the only wording this account is ever shown is the one the read route
+  // delivers for its language. The stored language is that one, decided
+  // here; a client naming another wording answered something this server
+  // did not offer (or the account language changed under an open form), and
+  // is refused rather than scored against the wrong table.
+  const offered = auditCInstrument(user.locale ?? "en").locale;
+  if (
+    input.locale !== undefined &&
+    auditCInstrument(input.locale).locale !== offered
+  ) {
+    return apiError("The wording answered is not the one offered", 422, {
+      errorCode: "anamnesis.auditC.invalid",
+    });
+  }
+  // An entry dated in the future would hold the "newest" place for as long
+  // as the clock takes to reach it. A few minutes of device clock skew pass.
+  const takenAt = input.takenAt ? new Date(input.takenAt) : undefined;
+  if (takenAt && takenAt.getTime() > Date.now() + AUDIT_C_CLOCK_SKEW_MS) {
+    return apiError("takenAt lies in the future", 422, {
+      errorCode: "anamnesis.auditC.invalid",
+    });
+  }
+
   const { result, replayed } = await createAuditC(user.id, {
     items: input.items,
-    takenAt: input.takenAt ? new Date(input.takenAt) : undefined,
+    takenAt,
     tz: input.tz,
-    locale: input.locale ?? user.locale ?? "en",
+    locale: offered,
     source: input.source,
     externalId: input.externalId,
   });
diff --git a/src/app/api/auth/me/modules/route.ts b/src/app/api/auth/me/modules/route.ts
index 5c10e8d93..abeceb8de 100644
--- a/src/app/api/auth/me/modules/route.ts
+++ b/src/app/api/auth/me/modules/route.ts
@@ -44,6 +44,7 @@ import {
 } from "@/lib/optimistic-lock";
 import { resolveModuleMap, normalisePrefs } from "@/lib/modules/gate";
 import { modulePrefsPatchSchema } from "@/lib/validations/modules";
+import { recomputeSupplementsIfNutrientsTurnedOn } from "@/lib/medications/ingredients/recompute-job";
 
 export const dynamic = "force-dynamic";
 
@@ -140,6 +141,12 @@ export const PATCH = apiHandler(async (req: Request) => {
   });
   if ("conflict" in guarded) return guarded.conflict;
 
+  await recomputeSupplementsIfNutrientsTurnedOn(
+    user.id,
+    existing?.modulePreferencesJson,
+    merged,
+  );
+
   // A module toggle can add or remove a Health Score pillar, so the
   // cached composite was computed from a composition that no longer
   // holds. Without this the score stays stale for up to an hour after a
diff --git a/src/app/api/auth/me/timezone/__tests__/route.test.ts b/src/app/api/auth/me/timezone/__tests__/route.test.ts
index 8fa2a9216..5cd3cfff8 100644
--- a/src/app/api/auth/me/timezone/__tests__/route.test.ts
+++ b/src/app/api/auth/me/timezone/__tests__/route.test.ts
@@ -31,6 +31,12 @@ vi.mock("@/lib/rollups/medication-compliance-rollups", () => ({
   enqueueUserMedicationComplianceBackfill: refoldMock,
 }));
 
+// The supplement nutrient rows are keyed by the same local day.
+const nutrientMock = vi.hoisted(() => vi.fn(async () => ({ queued: true })));
+vi.mock("@/lib/medications/ingredients/recompute-job", () => ({
+  enqueueSupplementNutrientRecompute: nutrientMock,
+}));
+
 vi.mock("next/headers", () => ({
   headers: vi.fn(async () => ({ get: () => null })),
   cookies: vi.fn(async () => ({
@@ -118,6 +124,7 @@ describe("PUT /api/auth/me/timezone", () => {
     const res = await PUT(mkReq({ timezone: "America/New_York" }));
     expect(res.status).toBe(200);
     expect(refoldMock).toHaveBeenCalledWith("user-1");
+    expect(nutrientMock).toHaveBeenCalledWith("user-1");
   });
 
   it("does not re-fold when the zone is unchanged", async () => {
@@ -130,6 +137,7 @@ describe("PUT /api/auth/me/timezone", () => {
     const res = await PUT(mkReq({ timezone: "Europe/Berlin" }));
     expect(res.status).toBe(200);
     expect(refoldMock).not.toHaveBeenCalled();
+    expect(nutrientMock).not.toHaveBeenCalled();
   });
 
   it("trims whitespace before validating", async () => {
diff --git a/src/app/api/auth/me/timezone/route.ts b/src/app/api/auth/me/timezone/route.ts
index 0bc0c0503..ad4ff0485 100644
--- a/src/app/api/auth/me/timezone/route.ts
+++ b/src/app/api/auth/me/timezone/route.ts
@@ -29,6 +29,7 @@ import {
 import { annotate } from "@/lib/logging/context";
 import { auditLog } from "@/lib/auth/audit";
 import { invalidateUserTimezone, isValidTimezone } from "@/lib/tz/resolver";
+import { enqueueSupplementNutrientRecompute } from "@/lib/medications/ingredients/recompute-job";
 import { enqueueUserMedicationComplianceBackfill } from "@/lib/rollups/medication-compliance-rollups";
 
 export const PUT = apiHandler(async (request: NextRequest) => {
@@ -69,6 +70,8 @@ export const PUT = apiHandler(async (request: NextRequest) => {
   // job; best-effort, the singleton key collapses rapid re-picks.
   if ((before?.timezone ?? null) !== tz) {
     const refold = await enqueueUserMedicationComplianceBackfill(user.id);
+    // The supplement nutrient rows are keyed by the same local day.
+    await enqueueSupplementNutrientRecompute(user.id);
     annotate({
       meta: {
         timezone_compliance_refold_enqueued: refold.enqueued,
diff --git a/src/app/api/auth/oidc/login/__tests__/route.test.ts b/src/app/api/auth/oidc/login/__tests__/route.test.ts
index 1d9fc1584..cf283fb6a 100644
--- a/src/app/api/auth/oidc/login/__tests__/route.test.ts
+++ b/src/app/api/auth/oidc/login/__tests__/route.test.ts
@@ -40,9 +40,25 @@ vi.mock("@/lib/auth/oidc", async () => {
   };
 });
 
+vi.mock("@/lib/managed-profiles/claim-cookie", async (importOriginal) => ({
+  ...(await importOriginal<
+    typeof import("@/lib/managed-profiles/claim-cookie")
+  >()),
+  readClaimCookie: vi.fn(),
+}));
+vi.mock("@/lib/managed-profiles/handover", () => ({
+  isHandoverUsable: vi.fn(),
+}));
+
 import { GET } from "../route";
+import { readClaimCookie } from "@/lib/managed-profiles/claim-cookie";
+import { isHandoverUsable } from "@/lib/managed-profiles/handover";
 import { checkAuthSurfaceRateLimit } from "@/lib/rate-limit";
-import { getOidcConfig, discoverOidcMetadata } from "@/lib/auth/oidc";
+import {
+  buildAuthorizationUrl,
+  discoverOidcMetadata,
+  getOidcConfig,
+} from "@/lib/auth/oidc";
 
 function makeRequest(path = "/api/auth/oidc/login"): NextRequest {
   return new NextRequest(`http://localhost${path}`);
@@ -248,6 +264,34 @@ describe("GET /api/auth/oidc/login?intent=claim (v1.43, #959)", () => {
     expect(res.cookies.get("oidc_auth_state")).toBeUndefined();
   });
 
+  it("asks the provider for a fresh sign-in on a claim, and only there", async () => {
+    vi.mocked(discoverOidcMetadata).mockResolvedValue({
+      issuer: "https://idp.example.com",
+      authorization_endpoint: "https://idp.example.com/authorize",
+      token_endpoint: "https://idp.example.com/token",
+      jwks_uri: "https://idp.example.com/jwks",
+    });
+    vi.mocked(buildAuthorizationUrl).mockReturnValue(
+      "https://idp.example.com/authorize?x=1",
+    );
+    vi.mocked(readClaimCookie).mockReturnValue({
+      handoverId: "handover-1",
+      binding: "binding-1",
+    } as never);
+    vi.mocked(isHandoverUsable).mockResolvedValue(true);
+
+    await GET(makeRequest("/api/auth/oidc/login?intent=claim"));
+    expect(vi.mocked(buildAuthorizationUrl).mock.calls[0]?.[0]).toMatchObject({
+      forceLogin: true,
+    });
+
+    vi.mocked(buildAuthorizationUrl).mockClear();
+    await GET(makeRequest("/api/auth/oidc/login"));
+    expect(vi.mocked(buildAuthorizationUrl).mock.calls[0]?.[0]).toMatchObject({
+      forceLogin: false,
+    });
+  });
+
   it("routes the claim's other failures to the claim page too", async () => {
     vi.mocked(checkAuthSurfaceRateLimit).mockResolvedValue({
       allowed: false,
diff --git a/src/app/api/auth/oidc/login/route.ts b/src/app/api/auth/oidc/login/route.ts
index dd0dc1cc8..f783e171c 100644
--- a/src/app/api/auth/oidc/login/route.ts
+++ b/src/app/api/auth/oidc/login/route.ts
@@ -156,6 +156,8 @@ export const GET = apiHandler(async (req: NextRequest) => {
     nonce,
     codeChallenge,
     redirectUri,
+    // A claim only; an ordinary sign-in keeps single sign-on.
+    forceLogin: claimBinding !== null,
   });
 
   const response = NextResponse.redirect(authorizationUrl);
diff --git a/src/app/api/coach/about-me/__tests__/route-concurrency.test.ts b/src/app/api/coach/about-me/__tests__/route-concurrency.test.ts
index 77954ad35..022796ebb 100644
--- a/src/app/api/coach/about-me/__tests__/route-concurrency.test.ts
+++ b/src/app/api/coach/about-me/__tests__/route-concurrency.test.ts
@@ -50,6 +50,7 @@ const { profile, userUpdate, transaction } = vi.hoisted(() => {
   };
   const userUpdate = vi.fn(async () => ({}));
   const tx = {
+    $queryRaw: vi.fn(async () => []),
     userHealthProfile: profile,
     user: { update: userUpdate },
   };
diff --git a/src/app/api/coach/about-me/__tests__/route-module-gate.test.ts b/src/app/api/coach/about-me/__tests__/route-module-gate.test.ts
index 402286c7d..95dcfe264 100644
--- a/src/app/api/coach/about-me/__tests__/route-module-gate.test.ts
+++ b/src/app/api/coach/about-me/__tests__/route-module-gate.test.ts
@@ -44,6 +44,7 @@ const {
   const profileUpsert = vi.fn(async () => ({ updatedAt: new Date(0) }));
   const userUpdate = vi.fn(async () => ({}));
   const tx = {
+    $queryRaw: vi.fn(async () => []),
     userHealthProfile: {
       findUnique: profileFindUnique,
       upsert: profileUpsert,
@@ -246,6 +247,7 @@ describe("about-me module gate", () => {
     transaction.mockImplementationOnce(
       async (callback: (client: unknown) => unknown) => {
         const result = await callback({
+          $queryRaw: vi.fn(async () => []),
           userHealthProfile: {
             findUnique: profileFindUnique,
             upsert: txProfileUpsert,
diff --git a/src/app/api/coach/about-me/route.ts b/src/app/api/coach/about-me/route.ts
index 5cc84ed6c..bacf26cad 100644
--- a/src/app/api/coach/about-me/route.ts
+++ b/src/app/api/coach/about-me/route.ts
@@ -42,6 +42,7 @@ import { annotate } from "@/lib/logging/context";
 import { auditLog } from "@/lib/auth/audit";
 import { invalidateUserInsights } from "@/lib/cache/invalidate";
 import { prisma } from "@/lib/db";
+import { lockAiSections } from "@/lib/profile/life-circumstances";
 import { takeBaseToken, invalidBaseTokenError } from "@/lib/optimistic-lock";
 import { checkRateLimit, rateLimitHeaders } from "@/lib/rate-limit";
 import { encryptToBytes } from "@/lib/ai/coach/bytes-codec";
@@ -246,6 +247,9 @@ export const PUT = apiHandler(async (req: Request) => {
     // A PUT here replaces the eight sections it lists and leaves the
     // life-circumstance switches (v1.43) exactly as stored.
     if (parsed.data.aiIncludedSections !== undefined) {
+      // Taken before the read, so a switch changed concurrently is read
+      // after it commits and carried over as it now stands.
+      await lockAiSections(tx, user.id);
       const stored = await tx.userHealthProfile.findUnique({
         where: { userId: user.id },
         select: { aiIncludedSections: true },
diff --git a/src/app/api/export/health-record/__tests__/route.test.ts b/src/app/api/export/health-record/__tests__/route.test.ts
index a141cccbf..3b61abb58 100644
--- a/src/app/api/export/health-record/__tests__/route.test.ts
+++ b/src/app/api/export/health-record/__tests__/route.test.ts
@@ -37,7 +37,7 @@ vi.mock("@/lib/db", () => ({
     userHealthProfile: { findUnique: vi.fn() },
     healthProfileFactRevision: { findMany: vi.fn() },
     // v1.43 — the anamnesis leaf also carries the newest AUDIT-C entry.
-    mentalHealthAssessment: { findFirst: vi.fn(async () => null) },
+    mentalHealthAssessment: { findMany: vi.fn(async () => []) },
     user: { findUnique: vi.fn(), update: vi.fn() },
     // The whole-record gate reads the session's own stamps: a sign-in a
     // moment ago is the proof, on an account without a second factor.
diff --git a/src/app/api/measurements/batch/route.ts b/src/app/api/measurements/batch/route.ts
index ea7c78dd1..876d7a75a 100644
--- a/src/app/api/measurements/batch/route.ts
+++ b/src/app/api/measurements/batch/route.ts
@@ -81,8 +81,10 @@ import {
 import {
   classifyHourBucket,
   HOUR_BUCKET_SKIP_REASONS,
+  combineHourWrites,
   parseHourBucketId,
   writeHourBuckets,
+  type HourBucketEntry,
   type HourBucketWrite,
 } from "@/lib/measurements/intraday-hour-buckets";
 import { resolveStoredTimezone } from "@/lib/tz/resolver";
@@ -394,7 +396,7 @@ async function postBatch(request: NextRequest): Promise<Response> {
   const unstableShapes: UnstableExternalIdShape[] = [];
   // v1.43 — hourly sums for the intraday profile, and the zone they are cut in
   // (the account's, the one `/api/auth/me` publishes as `timezone`).
-  const hourWrites: HourBucketWrite[] = [];
+  const hourWrites: HourBucketEntry[] = [];
   let accountTz: string | undefined;
   const batchNow = new Date();
 
@@ -484,6 +486,7 @@ async function postBatch(request: NextRequest): Promise<Response> {
         dateKey: hourId.dateKey,
         hour: hourId.hour,
         value: hourMapped.value,
+        startMs: startDate.getTime(),
       });
       continue;
     }
@@ -984,28 +987,24 @@ async function postBatch(request: NextRequest): Promise<Response> {
     }
   }
 
-  // v1.43 — the hours, after the measurements. The last entry for one slot
-  // wins, as for the day rows; an earlier one in the same batch is a
-  // superseded duplicate.
+  // v1.43 — the hours, after the measurements. One write per slot: the last
+  // entry for one start wins (an earlier one in the same batch is a
+  // superseded duplicate), and the two starts of the hour a fall-back day
+  // repeats are summed (`combineHourWrites`).
   let hourInsertedCount = 0;
   let hourUpdatedCount = 0;
   const hourLanded: HourBucketWrite[] = [];
   if (hourWrites.length > 0 && accountTz) {
-    const lastBySlot = new Map<string, HourBucketWrite>();
-    for (const w of hourWrites) {
-      lastBySlot.set(`${w.type}::${w.dateKey}::${w.hour}`, w);
+    const { writes, superseded } = combineHourWrites(hourWrites);
+    for (const index of superseded) {
+      results[index] = {
+        index,
+        status: "duplicate",
+        reason: "superseded_in_batch",
+      };
+      duplicateCount++;
     }
-    for (const w of hourWrites) {
-      if (lastBySlot.get(`${w.type}::${w.dateKey}::${w.hour}`) !== w) {
-        results[w.index] = {
-          index: w.index,
-          status: "duplicate",
-          reason: "superseded_in_batch",
-        };
-        duplicateCount++;
-      }
-    }
-    const toWrite = [...lastBySlot.values()];
+    const toWrite = writes.map(({ write }) => write);
     try {
       const statuses = await writeHourBuckets(
         prisma,
@@ -1013,21 +1012,25 @@ async function postBatch(request: NextRequest): Promise<Response> {
         accountTz,
         toWrite,
       );
-      for (const w of toWrite) {
-        const status = statuses.get(w.index) ?? "inserted";
-        results[w.index] = { index: w.index, status };
-        if (status === "inserted") hourInsertedCount++;
-        else hourUpdatedCount++;
-        hourLanded.push(w);
+      for (const { write, contributors } of writes) {
+        const status = statuses.get(write.index) ?? "inserted";
+        for (const index of contributors) {
+          results[index] = { index, status };
+          if (status === "inserted") hourInsertedCount++;
+          else hourUpdatedCount++;
+        }
+        hourLanded.push(write);
       }
     } catch (err) {
-      for (const w of toWrite) {
-        results[w.index] = {
-          index: w.index,
-          status: "failed",
-          reason: "write_failed",
-        };
-        failedCount++;
+      for (const { contributors } of writes) {
+        for (const index of contributors) {
+          results[index] = {
+            index,
+            status: "failed",
+            reason: "write_failed",
+          };
+          failedCount++;
+        }
       }
       annotate({
         action: { name: "measurement.batch.hour-bucket.failed" },
diff --git a/src/app/api/record-settings/[family]/route.ts b/src/app/api/record-settings/[family]/route.ts
index c899bd3ba..86b6b465c 100644
--- a/src/app/api/record-settings/[family]/route.ts
+++ b/src/app/api/record-settings/[family]/route.ts
@@ -1,5 +1,6 @@
 import type { NextRequest } from "next/server";
 
+import { enqueueSupplementNutrientRecompute } from "@/lib/medications/ingredients/recompute-job";
 import {
   apiHandler,
   requireGuardianAuth,
@@ -239,6 +240,8 @@ export const PATCH = apiHandler(
           }
           if (patch.timezone !== undefined) {
             invalidateUserMedications(access.recordId, { evict: true });
+            // The supplement nutrient rows are keyed by the local day.
+            await enqueueSupplementNutrientRecompute(access.recordId);
           }
           settings = {
             ...record,
diff --git a/src/components/admin/__tests__/pulse-day-mean-card.test.tsx b/src/components/admin/__tests__/pulse-day-mean-card.test.tsx
index 926de7ea6..317901ea6 100644
--- a/src/components/admin/__tests__/pulse-day-mean-card.test.tsx
+++ b/src/components/admin/__tests__/pulse-day-mean-card.test.tsx
@@ -37,7 +37,12 @@ const RESULT = {
 const status = (
   lastDryRun: PulseDayMeanStatus["lastDryRun"],
   method: PulseDayMeanStatus["method"] = "hour_mean_v1",
-): PulseDayMeanStatus => ({ method, appliedAt: null, lastDryRun });
+): PulseDayMeanStatus => ({
+  method,
+  appliedAt: null,
+  lastDryRun,
+  lastRebuild: null,
+});
 
 const done = (finishedAt: string) => ({
   status: "done" as const,
@@ -113,4 +118,33 @@ describe("PulseDayMeanCard", () => {
     expect(de).toContain("Dünne Stunden gewichtet");
     expect(de).toContain("Probelauf");
   });
+
+  it("shows a stopped recalculation and offers it again without a dry run", () => {
+    const failed: PulseDayMeanStatus = {
+      ...status(done("2026-10-01T10:00:00Z"), "weighted_hour_mean_v2"),
+      appliedAt: "2026-10-01T11:00:00Z",
+      lastRebuild: {
+        status: "failed",
+        startedAt: "2026-10-01T11:00:00Z",
+        finishedAt: "2026-10-01T11:05:00Z",
+        accounts: null,
+        rows: null,
+      },
+    };
+    const html = render(failed);
+    expect(html).toContain('data-state="failed"');
+    expect(html).toContain('data-testid="admin-pulse-day-mean-rebuild"');
+
+    const finished = render({
+      ...failed,
+      lastRebuild: {
+        ...failed.lastRebuild!,
+        status: "done",
+        accounts: 3,
+        rows: 900,
+      },
+    });
+    expect(finished).not.toContain("admin-pulse-day-mean-rebuild");
+    expect(finished).toContain('data-slot="pulse-day-mean-rebuild"');
+  });
 });
diff --git a/src/components/admin/pulse-day-mean-card.tsx b/src/components/admin/pulse-day-mean-card.tsx
index 285fec109..9a78004f4 100644
--- a/src/components/admin/pulse-day-mean-card.tsx
+++ b/src/components/admin/pulse-day-mean-card.tsx
@@ -57,8 +57,13 @@ export function PulseDayMeanCard() {
     queryFn: () =>
       apiGet<PulseDayMeanStatus>(ENDPOINT, { credentials: "include" }),
     refetchInterval: (query) => {
-      const state = query.state.data?.lastDryRun?.status;
-      return state === "queued" || state === "running" ? POLL_MS : false;
+      const states = [
+        query.state.data?.lastDryRun?.status,
+        query.state.data?.lastRebuild?.status,
+      ];
+      return states.some((state) => state === "queued" || state === "running")
+        ? POLL_MS
+        : false;
     },
   });
 
@@ -102,7 +107,11 @@ export function PulseDayMeanCard() {
             : t("admin.section.pulseDayMean.appliedNoRebuild"),
         );
       } else {
-        toast.success(t("admin.section.pulseDayMean.dryRunStarted"));
+        toast.success(
+          action === "rebuild"
+            ? t("admin.section.pulseDayMean.rebuildStarted")
+            : t("admin.section.pulseDayMean.dryRunStarted"),
+        );
       }
       void queryClient.invalidateQueries({
         queryKey: queryKeys.adminPulseDayMean(),
@@ -125,9 +134,16 @@ export function PulseDayMeanCard() {
   const s = status.data;
   const dryRun = s?.lastDryRun ?? null;
   const result = dryRun?.status === "done" ? dryRun.result : null;
-  const active = dryRun?.status === "queued" || dryRun?.status === "running";
+  const rebuild = s?.lastRebuild ?? null;
+  const rebuilding =
+    rebuild?.status === "queued" || rebuild?.status === "running";
+  const active =
+    dryRun?.status === "queued" || dryRun?.status === "running" || rebuilding;
   const weighted = s?.method === "weighted_hour_mean_v2";
   const canApply = !!s && !active && hasFreshDryRun(s);
+  // A stopped recalculation is run again from here; the method is already
+  // switched, so it needs no new dry run.
+  const canRebuild = !!s?.appliedAt && rebuild?.status === "failed";
 
   return (
     <SettingsCard data-testid="admin-pulse-day-mean">
@@ -252,6 +268,25 @@ export function PulseDayMeanCard() {
         </p>
       ) : null}
 
+      {rebuild ? (
+        <p
+          className={
+            rebuild.status === "failed"
+              ? "text-destructive text-sm"
+              : "text-muted-foreground text-xs"
+          }
+          data-slot="pulse-day-mean-rebuild"
+          data-state={rebuild.status}
+        >
+          {rebuild.status === "done"
+            ? t("admin.section.pulseDayMean.rebuild.done", {
+                at: fmt.dateTime(rebuild.finishedAt ?? rebuild.startedAt),
+                accounts: fmt.number(rebuild.accounts ?? 0),
+              })
+            : t(`admin.section.pulseDayMean.rebuild.${rebuild.status}`)}
+        </p>
+      ) : null}
+
       <SettingsCardActions>
         <Button
           size="sm"
@@ -264,6 +299,18 @@ export function PulseDayMeanCard() {
           <FlaskConical className="size-3.5" aria-hidden />
           {t("admin.section.pulseDayMean.dryRun")}
         </Button>
+        {canRebuild ? (
+          <Button
+            size="sm"
+            variant="outline"
+            className="min-h-11 sm:min-h-9"
+            disabled={run.isPending}
+            onClick={() => run.mutate("rebuild")}
+            data-testid="admin-pulse-day-mean-rebuild"
+          >
+            {t("admin.section.pulseDayMean.rebuildAgain")}
+          </Button>
+        ) : null}
         <AlertDialog>
           <AlertDialogTrigger asChild>
             <Button
diff --git a/src/components/records/__tests__/audit-c-legacy-pattern.test.tsx b/src/components/records/__tests__/audit-c-legacy-pattern.test.tsx
new file mode 100644
index 000000000..89a74a2a8
--- /dev/null
+++ b/src/components/records/__tests__/audit-c-legacy-pattern.test.tsx
@@ -0,0 +1,73 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+import { renderToStaticMarkup } from "react-dom/server";
+
+const state = vi.hoisted(() => ({
+  facts: null as unknown,
+}));
+
+vi.mock("@tanstack/react-query", () => ({
+  useQuery: () => ({ data: state.facts }),
+  useQueryClient: () => ({ invalidateQueries: vi.fn() }),
+  useMutation: () => ({ mutate: vi.fn(), isPending: false }),
+}));
+vi.mock("@/lib/api/api-fetch", () => ({
+  apiGet: vi.fn(),
+  apiPost: vi.fn(),
+  apiDelete: vi.fn(),
+}));
+vi.mock("@/components/outcome/outcome-toast", () => ({
+  toastWrittenOutcome: vi.fn(),
+}));
+
+import { I18nProvider } from "@/lib/i18n/context";
+import { auditCInstrument } from "@/lib/mental-health/audit-c";
+import { AuditCPanel } from "../audit-c-panel";
+
+function render() {
+  return renderToStaticMarkup(
+    <I18nProvider initialLocale="en">
+      <AuditCPanel
+        data={{
+          latest: null,
+          history: [],
+          offerDue: true,
+          instrument: auditCInstrument("en"),
+        }}
+        onChanged={() => {}}
+      />
+    </I18nProvider>,
+  );
+}
+
+beforeEach(() => {
+  state.facts = null;
+});
+
+describe("the old alcohol pattern in the alcohol section", () => {
+  it("lists the stored pattern with a way to remove it", () => {
+    state.facts = {
+      current: {
+        SMOKING_STATUS: null,
+        SHIFT_SCHEDULE: null,
+        ALCOHOL_PATTERN: { id: "fact-1", value: "WEEKLY" },
+      },
+      history: [],
+    };
+    const html = render();
+    expect(html).toContain('data-slot="audit-c-legacy-pattern"');
+    expect(html).toContain("Earlier answer: Weekly");
+    expect(html).toContain("audit-c-legacy-remove");
+  });
+
+  it("shows nothing when there is no stored pattern", () => {
+    state.facts = {
+      current: {
+        SMOKING_STATUS: null,
+        SHIFT_SCHEDULE: null,
+        ALCOHOL_PATTERN: null,
+      },
+      history: [],
+    };
+    expect(render()).not.toContain("audit-c-legacy-pattern");
+  });
+});
diff --git a/src/components/records/audit-c-panel.tsx b/src/components/records/audit-c-panel.tsx
index 233d51e6e..4edc2a06c 100644
--- a/src/components/records/audit-c-panel.tsx
+++ b/src/components/records/audit-c-panel.tsx
@@ -10,7 +10,7 @@
  * and its source are the server's too; nothing is scored here.
  */
 import { useState } from "react";
-import { useMutation } from "@tanstack/react-query";
+import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
 import { Loader2, Trash2 } from "lucide-react";
 
 import { toastWrittenOutcome } from "@/components/outcome/outcome-toast";
@@ -18,9 +18,10 @@ import { SettingsCardActions } from "@/components/settings/_card-actions";
 import { Button } from "@/components/ui/button";
 import { ConfirmButton } from "@/components/ui/confirm-button";
 import { ResponsiveSheet } from "@/components/ui/responsive-sheet";
-import { apiDelete, apiPost } from "@/lib/api/api-fetch";
+import { apiDelete, apiGet, apiPost } from "@/lib/api/api-fetch";
 import { showErrorToast } from "@/lib/errors/error-toast";
 import { useFormatters, useTranslations } from "@/lib/i18n/context";
+import type { HealthProfileFactsPayload } from "@/lib/profile/health-facts";
 import type { LifeCircumstancesDto } from "@/lib/profile/life-circumstances";
 import { queryKeys } from "@/lib/query-keys";
 
@@ -76,6 +77,34 @@ export function AuditCPanel({
 
   const complete = answers.every((answer) => answer !== null);
 
+  // The one-word alcohol pattern asked before AUDIT-C. It still lives in the
+  // record, so it is shown here, where alcohol now lives, with a way out.
+  // The Coach stops reading it once an AUDIT-C entry exists.
+  const queryClient = useQueryClient();
+  const facts = useQuery({
+    queryKey: queryKeys.healthProfileFacts(),
+    queryFn: () => apiGet<HealthProfileFactsPayload>("/api/anamnesis/facts"),
+  });
+  const legacy = facts.data?.current.ALCOHOL_PATTERN ?? null;
+  const removeLegacy = useMutation({
+    mutationKey: queryKeys.healthProfileFacts(),
+    mutationFn: (id: string) => apiDelete(`/api/anamnesis/facts/${id}`),
+    onSuccess: () => {
+      toastWrittenOutcome(
+        "success",
+        t("records.life.auditC.legacyRemovedToast"),
+      );
+      void queryClient.invalidateQueries({
+        queryKey: queryKeys.healthProfileFacts(),
+      });
+      onChanged();
+    },
+    onError: (error) =>
+      showErrorToast(t("records.life.auditC.legacyRemoveError"), {
+        cause: error,
+      }),
+  });
+
   return (
     <div className="space-y-4" data-slot="audit-c-panel">
       <p className="text-sm">{t("records.life.auditC.intro")}</p>
@@ -154,6 +183,41 @@ export function AuditCPanel({
         </div>
       ) : null}
 
+      {legacy?.value ? (
+        <div
+          className="flex items-start justify-between gap-2"
+          data-slot="audit-c-legacy-pattern"
+        >
+          <div className="space-y-1">
+            <p className="text-sm">
+              {t("records.life.auditC.legacyLabel", {
+                value: t(
+                  `records.profileFacts.values.ALCOHOL_PATTERN.${legacy.value}`,
+                ),
+              })}
+            </p>
+            <p className="text-muted-foreground text-xs">
+              {t("records.life.auditC.legacyHint")}
+            </p>
+          </div>
+          <ConfirmButton
+            label={t("records.life.auditC.legacyRemove")}
+            ariaLabel={t("records.life.auditC.legacyRemove")}
+            title={t("records.life.auditC.legacyRemoveConfirmTitle")}
+            body={t("records.life.auditC.legacyRemoveConfirmBody")}
+            confirmLabel={t("records.life.auditC.legacyRemove")}
+            onConfirm={() => removeLegacy.mutate(legacy.id)}
+            pending={removeLegacy.isPending}
+            disabled={removeLegacy.isPending}
+            variant="ghost"
+            size="icon"
+            className="text-destructive min-h-11 sm:min-h-9"
+            icon={<Trash2 className="size-4" aria-hidden />}
+            slot="audit-c-legacy-remove"
+          />
+        </div>
+      ) : null}
+
       {data.offerDue ? (
         <p className="text-muted-foreground text-xs">
           {t("records.life.auditC.offerDue")}
diff --git a/src/lib/ai/__tests__/usage-level.test.ts b/src/lib/ai/__tests__/usage-level.test.ts
index 7b06d9c0c..bee432add 100644
--- a/src/lib/ai/__tests__/usage-level.test.ts
+++ b/src/lib/ai/__tests__/usage-level.test.ts
@@ -61,6 +61,9 @@ describe("resolveAiUsageLevel — every choice x default x ceiling", () => {
 describe("what each level does", () => {
   it("regenerates status cards after a sync: a day, the operator's gap, the hour", () => {
     expect(statusRegenGapHours("frugal", "sync", 6)).toBe(24);
+    // Never more often than standard: an operator gap above a day wins.
+    expect(statusRegenGapHours("frugal", "sync", 48)).toBe(48);
+    expect(statusRegenGapHours("standard", "sync", 48)).toBe(48);
     expect(statusRegenGapHours("standard", "sync", 6)).toBe(6);
     expect(statusRegenGapHours("standard", "sync", 12)).toBe(12);
     expect(statusRegenGapHours("full", "sync", 12)).toBe(1);
diff --git a/src/lib/ai/coach/__tests__/profile-ai-section-gating.test.ts b/src/lib/ai/coach/__tests__/profile-ai-section-gating.test.ts
index 2a3b575c1..e9c7dabf4 100644
--- a/src/lib/ai/coach/__tests__/profile-ai-section-gating.test.ts
+++ b/src/lib/ai/coach/__tests__/profile-ai-section-gating.test.ts
@@ -6,7 +6,7 @@ const mocks = vi.hoisted(() => ({
   allergyFindMany: vi.fn(),
   familyFindMany: vi.fn(),
   factFindMany: vi.fn(),
-  auditFindFirst: vi.fn(),
+  auditFindMany: vi.fn(),
   decrypt: vi.fn(),
 }));
 
@@ -17,7 +17,7 @@ vi.mock("@/lib/db", () => ({
     allergy: { findMany: mocks.allergyFindMany },
     familyHistoryEntry: { findMany: mocks.familyFindMany },
     healthProfileFactRevision: { findMany: mocks.factFindMany },
-    mentalHealthAssessment: { findFirst: mocks.auditFindFirst },
+    mentalHealthAssessment: { findMany: mocks.auditFindMany },
   },
 }));
 
@@ -46,13 +46,15 @@ beforeEach(() => {
   mocks.allergyFindMany.mockResolvedValue([]);
   mocks.familyFindMany.mockResolvedValue([]);
   mocks.factFindMany.mockResolvedValue([]);
-  mocks.auditFindFirst.mockResolvedValue(null);
+  mocks.auditFindMany.mockResolvedValue([]);
   mocks.decrypt.mockImplementation((bytes: Uint8Array) => {
     if (bytes[0] === 1) return secret;
     if (bytes[0] === 2) return "Asthma";
     if (bytes[0] === 3) return "FORMER";
     if (bytes[0] === 4) return "OCCASIONAL";
     if (bytes[0] === 5) return "ROTATING";
+    if (bytes[0] === 6) return JSON.stringify({ schema: 1, items: [2, 2, 1] });
+    if (bytes[0] === 7) throw new Error("unknown key id");
     return "unknown";
   });
 });
@@ -204,6 +206,11 @@ describe("profile section access in AI prompt paths", () => {
         if (bytes[0] === 21) {
           return JSON.stringify({ schema: 1, goals: ["sleep"] });
         }
+        if (bytes[0] === 4) return "OCCASIONAL";
+        if (bytes[0] === 6) {
+          return JSON.stringify({ schema: 1, items: [2, 2, 1] });
+        }
+        if (bytes[0] === 7) throw new Error("unknown key id");
         return "unknown";
       });
     });
@@ -235,11 +242,16 @@ describe("profile section access in AI prompt paths", () => {
     });
 
     it("gives AUDIT-C as a sum with its screening verdict under the alcohol switch only", async () => {
-      mocks.auditFindFirst.mockResolvedValue({
-        totalScore: 5,
-        takenAt: new Date("2026-10-01T09:00:00.000Z"),
-        locale: "de",
-      });
+      mocks.auditFindMany.mockResolvedValue([
+        {
+          id: "audit-1",
+          totalScore: 5,
+          takenAt: new Date("2026-10-01T09:00:00.000Z"),
+          locale: "de",
+          tz: null,
+          responsesEncrypted: new Uint8Array([6]),
+        },
+      ]);
       mocks.userFindUnique.mockResolvedValue({ gender: "MALE", locale: "de" });
 
       mockIncluded(["ALCOHOL_PATTERN"]);
@@ -248,10 +260,59 @@ describe("profile section access in AI prompt paths", () => {
         "AUDIT-C (Screening, keine Diagnose): 5 von 12 am 2026-10-01, Screening-Grenzwert 5 erreicht",
       );
 
-      mocks.auditFindFirst.mockClear();
+      mocks.auditFindMany.mockClear();
       mockIncluded(["GOALS"]);
       await getSelfContextTextForUser("user-1", "de");
-      expect(mocks.auditFindFirst).not.toHaveBeenCalled();
+      expect(mocks.auditFindMany).not.toHaveBeenCalled();
+    });
+
+    it("drops the old alcohol pattern once an AUDIT-C entry exists", async () => {
+      mockIncluded(["ALCOHOL_PATTERN"]);
+      mocks.factFindMany.mockResolvedValue([
+        { kind: "ALCOHOL_PATTERN", valueEncrypted: new Uint8Array([4]) },
+      ]);
+      const alone = await getSelfContextTextForUser("user-1", "en");
+      expect(alone).toContain("Alcohol pattern: Occasional");
+
+      mocks.auditFindMany.mockResolvedValue([
+        {
+          id: "audit-1",
+          totalScore: 5,
+          takenAt: new Date("2026-10-01T09:00:00.000Z"),
+          locale: "en",
+          tz: null,
+          responsesEncrypted: new Uint8Array([6]),
+        },
+      ]);
+      const text = await getSelfContextTextForUser("user-1", "en");
+      expect(text).toContain("AUDIT-C");
+      expect(text).not.toContain("Alcohol pattern");
+    });
+
+    it("reads the newest entry whose answers open, as the read route does", async () => {
+      mocks.userFindUnique.mockResolvedValue({ gender: null, locale: "en" });
+      mocks.auditFindMany.mockResolvedValue([
+        {
+          id: "audit-sealed",
+          totalScore: 11,
+          takenAt: new Date("2026-10-05T09:00:00.000Z"),
+          locale: "en",
+          tz: null,
+          responsesEncrypted: new Uint8Array([7]),
+        },
+        {
+          id: "audit-open",
+          totalScore: 5,
+          takenAt: new Date("2026-10-01T09:00:00.000Z"),
+          locale: "en",
+          tz: null,
+          responsesEncrypted: new Uint8Array([6]),
+        },
+      ]);
+      mockIncluded(["ALCOHOL_PATTERN"]);
+      const text = await getSelfContextTextForUser("user-1", "en");
+      expect(text).toContain("5 of 12");
+      expect(text).not.toContain("11 of 12");
     });
   });
 });
diff --git a/src/lib/ai/coach/about-me.ts b/src/lib/ai/coach/about-me.ts
index 4b5423721..6d251e4ad 100644
--- a/src/lib/ai/coach/about-me.ts
+++ b/src/lib/ai/coach/about-me.ts
@@ -47,6 +47,7 @@ import {
 } from "@/lib/profile/health-facts";
 import { LIFE_KIND_AI_SECTION } from "@/lib/profile/life-derived";
 import { auditCCutoff } from "@/lib/mental-health/audit-c";
+import { loadNewestReadableAuditC } from "@/lib/profile/audit-c-entries";
 import {
   LIFE_FACT_KINDS,
   isLifeFactKind,
@@ -636,20 +637,16 @@ export function composeLifeCircumstancesText(
 
 /**
  * The newest AUDIT-C entry for a prompt: only when the alcohol section is
- * released, and without decrypting anything (the sum is stored in the clear,
- * like the screeners'). The profile sex for the cut-off is read only when
- * there is an entry to judge.
+ * released. "Newest" is the newest entry whose answers open, the same entry
+ * the read route shows as the result (`audit-c-entries.ts`). The profile sex
+ * for the cut-off is read only when there is an entry to judge.
  */
 async function loadAuditCForPrompt(
   userId: string,
   included: ReadonlySet<string>,
 ): Promise<AuditCForPrompt | null> {
   if (!included.has("ALCOHOL_PATTERN")) return null;
-  const audit = await prisma.mentalHealthAssessment.findFirst({
-    where: { userId, instrument: "AUDIT_C", deletedAt: null },
-    orderBy: [{ takenAt: "desc" }, { id: "desc" }],
-    select: { totalScore: true, takenAt: true, locale: true, tz: true },
-  });
+  const audit = await loadNewestReadableAuditC(userId);
   if (!audit) return null;
   const user = await prisma.user.findUnique({
     where: { id: userId },
@@ -839,6 +836,10 @@ export async function getSelfContextTextForUser(
     const factsText = composeStructuredHealthFactsText(
       factRows
         .filter((row) => !isLifeFactKind(row.kind))
+        // The old one-word alcohol pattern gives way to an AUDIT-C entry:
+        // the questionnaire is the measured answer to the same question, and
+        // the two side by side would let a years-old guess argue with it.
+        .filter((row) => !(auditC && row.kind === "ALCOHOL_PATTERN"))
         .flatMap((row) => {
           const decrypted = decryptHealthProfileFactValue(
             row.kind as HealthProfileFactKind,
diff --git a/src/lib/ai/coach/snapshot.ts b/src/lib/ai/coach/snapshot.ts
index 4c2a53dd8..2b2a4a01e 100644
--- a/src/lib/ai/coach/snapshot.ts
+++ b/src/lib/ai/coach/snapshot.ts
@@ -15,7 +15,7 @@
  * values with weekday labels and aggregates the older window into
  * weekly buckets so the prompt budget stays tight.
  */
-import { loadFreeWeekdaySpans } from "@/lib/profile/life-circumstances";
+import { loadFreeWeekdaySpansForPrompt } from "@/lib/profile/life-circumstances";
 import { prisma } from "@/lib/db";
 import { extractFeatures } from "@/lib/insights/features";
 import type { CoachDataCluster } from "@/lib/validations/coach-prefs";
@@ -807,7 +807,7 @@ async function buildCoachSnapshotImpl(
       sourcePriorityJson: prefsRow?.sourcePriorityJson ?? null,
       userTz,
       ageYears: derivedProfile.ageYears,
-      freeWeekdaySpans: await loadFreeWeekdaySpans(userId),
+      freeWeekdaySpans: await loadFreeWeekdaySpansForPrompt(userId),
       snapshot,
       metrics,
       registerBlock,
diff --git a/src/lib/ai/usage-level.ts b/src/lib/ai/usage-level.ts
index e81673b15..cba27e992 100644
--- a/src/lib/ai/usage-level.ts
+++ b/src/lib/ai/usage-level.ts
@@ -44,9 +44,12 @@ export interface AiUsageLevelEffect {
    * Minimum hours between two status-card regenerations of one scope after a
    * sync (a watch, an integration). `"instance"` reads the operator's sync
    * gap (`AppSettings.aiSyncRegenGapHours`, else `AI_SYNC_REGEN_GAP_HOURS`,
-   * else six).
+   * else six). `{ instanceAtLeast }` reads the same gap but never goes
+   * below the floor, so a level meant to regenerate less often than
+   * `standard` cannot end up regenerating more often on an instance whose
+   * operator set a gap above the floor.
    */
-  syncRegenGapHours: number | "instance";
+  syncRegenGapHours: number | "instance" | { instanceAtLeast: number };
   /** The same gap after a hand entry. */
   manualRegenGapHours: number;
   /**
@@ -89,7 +92,7 @@ export const AI_USAGE_LEVEL_EFFECTS: Readonly<
   Record<AiUsageLevel, Readonly<AiUsageLevelEffect>>
 > = Object.freeze({
   frugal: {
-    syncRegenGapHours: 24,
+    syncRegenGapHours: { instanceAtLeast: 24 },
     manualRegenGapHours: 6,
     statusSnapshotChars: 10_000,
     backgroundReasoning: false,
@@ -235,9 +238,10 @@ export function statusRegenGapHours(
 ): number {
   const effect = AI_USAGE_LEVEL_EFFECTS[level];
   if (source === "manual") return effect.manualRegenGapHours;
-  return effect.syncRegenGapHours === "instance"
-    ? instanceSyncGapHours
-    : effect.syncRegenGapHours;
+  const gap = effect.syncRegenGapHours;
+  if (gap === "instance") return instanceSyncGapHours;
+  if (typeof gap === "number") return gap;
+  return Math.max(gap.instanceAtLeast, instanceSyncGapHours);
 }
 
 // ── Estimate ────────────────────────────────────────────────────────────────
diff --git a/src/lib/auth/__tests__/oidc.test.ts b/src/lib/auth/__tests__/oidc.test.ts
index dbcadf1a1..d2f821cdc 100644
--- a/src/lib/auth/__tests__/oidc.test.ts
+++ b/src/lib/auth/__tests__/oidc.test.ts
@@ -253,6 +253,23 @@ describe("buildAuthorizationUrl", () => {
     expect(url.searchParams.get("code_challenge")).toBe("challenge-1");
     expect(url.searchParams.get("code_challenge_method")).toBe("S256");
     expect(url.searchParams.get("scope")).toBe(config.scopes);
+    expect(url.searchParams.has("prompt")).toBe(false);
+  });
+
+  it("asks for a fresh sign-in only when told to", () => {
+    configureEnv();
+    const url = new URL(
+      buildAuthorizationUrl({
+        metadata: METADATA_DOC,
+        config: getOidcConfig()!,
+        state: "state-1",
+        nonce: "nonce-1",
+        codeChallenge: "challenge-1",
+        redirectUri: "https://healthlog.example.com/api/auth/oidc/callback",
+        forceLogin: true,
+      }),
+    );
+    expect(url.searchParams.get("prompt")).toBe("login");
   });
 });
 
diff --git a/src/lib/auth/oidc.ts b/src/lib/auth/oidc.ts
index 3282bd95e..1c5b86f56 100644
--- a/src/lib/auth/oidc.ts
+++ b/src/lib/auth/oidc.ts
@@ -237,6 +237,13 @@ export function buildAuthorizationUrl(params: {
   nonce: string;
   codeChallenge: string;
   redirectUri: string;
+  /**
+   * Ask the provider to authenticate afresh (`prompt=login`, OpenID Connect
+   * Core 3.1.2.1) instead of reusing the browser's provider session. A claim
+   * hands a profile to whoever signs in, so it must be the person at the
+   * keyboard now, not a session someone else left open in this browser.
+   */
+  forceLogin?: boolean;
 }): string {
   const url = new URL(params.metadata.authorization_endpoint);
   url.search = new URLSearchParams({
@@ -248,6 +255,7 @@ export function buildAuthorizationUrl(params: {
     nonce: params.nonce,
     code_challenge: params.codeChallenge,
     code_challenge_method: "S256",
+    ...(params.forceLogin ? { prompt: "login" } : {}),
   }).toString();
   return url.toString();
 }
diff --git a/src/lib/dawarich/__tests__/client-body-cap.test.ts b/src/lib/dawarich/__tests__/client-body-cap.test.ts
new file mode 100644
index 000000000..19c9fe2e4
--- /dev/null
+++ b/src/lib/dawarich/__tests__/client-body-cap.test.ts
@@ -0,0 +1,42 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+const safeFetchMock = vi.hoisted(() => vi.fn());
+vi.mock("@/lib/safe-fetch", async (importOriginal) => ({
+  ...(await importOriginal<typeof import("@/lib/safe-fetch")>()),
+  safeFetch: safeFetchMock,
+}));
+
+import { DawarichApiError, probeDawarich } from "../client";
+
+const CONN = { baseUrl: "https://dawarich.example.com", apiKey: "key" };
+
+beforeEach(() => {
+  safeFetchMock.mockReset();
+});
+
+describe("reading a Dawarich answer", () => {
+  it("parses an honest answer", async () => {
+    safeFetchMock.mockResolvedValue(
+      new Response(JSON.stringify({ user: { id: 1 } }), { status: 200 }),
+    );
+    await expect(probeDawarich(CONN)).resolves.toBeUndefined();
+  });
+
+  it("stops reading an answer that runs past the cap", async () => {
+    // Streamed without a length, so only the running count can catch it.
+    const chunk = new Uint8Array(1024 * 1024).fill(32);
+    let sent = 0;
+    const body = new ReadableStream<Uint8Array>({
+      pull(controller) {
+        sent += 1;
+        if (sent > 64) controller.close();
+        else controller.enqueue(chunk);
+      },
+    });
+    safeFetchMock.mockResolvedValue(new Response(body, { status: 200 }));
+    const failure = await probeDawarich(CONN).catch((e: unknown) => e);
+    expect(failure).toBeInstanceOf(DawarichApiError);
+    expect((failure as Error).message).toMatch(/oversized/);
+    expect(sent).toBeLessThan(64);
+  });
+});
diff --git a/src/lib/dawarich/client.ts b/src/lib/dawarich/client.ts
index e90f5d06b..68fe4e62b 100644
--- a/src/lib/dawarich/client.ts
+++ b/src/lib/dawarich/client.ts
@@ -5,7 +5,9 @@
  * reads four endpoints, all with `Authorization: Bearer <api key>`:
  *
  *   GET /api/v1/timeline?start_at&end_at&distance_unit=km
- *       one entry per local day (the instance's own zone) with a `summary`:
+ *       one entry per day in the Dawarich account's own zone (its timezone
+ *       setting, the instance's `TIME_ZONE` by default; the request cannot
+ *       name one), summing only what lies inside the window, with a `summary`:
  *       `total_distance`, `places_visited`, `mode_distances.{driving,train}`
  *       in km to one decimal. At most 31 days per request.
  *   GET /api/v1/visits?start_at&end_at
@@ -26,6 +28,7 @@
  * `DAWARICH_PRIVATE_ORIGINS`. The policy is re-read on every call, so taking
  * an origin off the list stops the next request.
  */
+import { readBodyCapped } from "@/lib/http/read-capped";
 import { SafeFetchError, safeFetch } from "@/lib/safe-fetch";
 
 import {
@@ -35,6 +38,14 @@ import {
 
 const DEFAULT_TIMEOUT_MS = 20_000;
 
+/**
+ * The most any one answer may hold. A month of timeline with its entries and
+ * a year of residency stay far below it; the cap is what keeps an instance
+ * that streams without end from holding the worker's memory until the
+ * timeout.
+ */
+const MAX_BODY_BYTES = 16 * 1024 * 1024;
+
 /** Dawarich refuses a timeline window longer than 31 days. */
 export const TIMELINE_MAX_DAYS = 31;
 
@@ -118,8 +129,12 @@ async function dawarichGet(
   if (res.status < 200 || res.status >= 300) {
     throw new DawarichApiError(`Dawarich responded ${res.status}`, res.status);
   }
+  const read = await readBodyCapped(res, MAX_BODY_BYTES);
+  if (!read.ok) {
+    throw new DawarichApiError("Dawarich returned an oversized body", null);
+  }
   try {
-    return await res.json();
+    return JSON.parse(read.text) as unknown;
   } catch {
     throw new DawarichApiError("Dawarich returned a non-JSON body", null);
   }
diff --git a/src/lib/dawarich/sync.ts b/src/lib/dawarich/sync.ts
index b1d6712ef..98f4e458e 100644
--- a/src/lib/dawarich/sync.ts
+++ b/src/lib/dawarich/sync.ts
@@ -46,8 +46,57 @@ const HOME_COUNTRY_WINDOW_DAYS = 365;
 /** How far before the range the visits request reaches. */
 const VISIT_LOOKBACK_MS = 7 * 24 * 60 * 60 * 1000;
 
-/** Days per timeline request, inside Dawarich's 31-day limit. */
-const TIMELINE_CHUNK_DAYS = TIMELINE_MAX_DAYS - 1;
+/**
+ * Days kept per timeline request. Each request reaches one whole day past
+ * both ends of its chunk (see `timelineChunks`), so it spans the chunk plus
+ * two days, and a fall-back day is 25 hours long: chunk + 2 must stay under
+ * Dawarich's 31-day limit by more than an hour.
+ */
+const TIMELINE_CHUNK_DAYS = TIMELINE_MAX_DAYS - 3;
+
+/**
+ * The timeline requests for `days`, each with the dates it is allowed to
+ * fill.
+ *
+ * Dawarich groups a timeline by the day in ITS user's zone (the timezone
+ * setting of the Dawarich account, `TIME_ZONE` by default), not by the zone
+ * the request names, and it only sums what lies inside the requested window.
+ * A window cut at the account's local midnight therefore answers the edge day
+ * in the instance's zone with a fragment of it whenever the two zones differ.
+ * Two rules keep a fragment out of the stored values:
+ *
+ *   - each request reaches a whole day before and after its chunk, so every
+ *     date of the chunk is covered end to end in any zone within a day of the
+ *     account's;
+ *   - only the chunk's own dates are kept from the answer, so the padded edge
+ *     days (the fragments) are thrown away and no chunk can overwrite a date
+ *     another chunk owns.
+ *
+ * The API offers no zone parameter and no per-track detail to regroup by, so
+ * the stored value for a date is the instance's day of that date. When the
+ * two zones agree, as on a typical self-host, that is exactly the account's
+ * local day.
+ */
+function timelineChunks(
+  days: readonly string[],
+  tz: string,
+): Array<{ startAt: Date; endAt: Date; dates: Set<string> }> {
+  const out: Array<{ startAt: Date; endAt: Date; dates: Set<string> }> = [];
+  for (let i = 0; i < days.length; i += TIMELINE_CHUNK_DAYS) {
+    const chunk = days.slice(i, i + TIMELINE_CHUNK_DAYS);
+    const { dayStart } = localDayWindow(shiftDateKey(chunk[0], -1), tz);
+    const { dayEnd } = localDayWindow(
+      shiftDateKey(chunk[chunk.length - 1], 1),
+      tz,
+    );
+    out.push({
+      startAt: dayStart,
+      endAt: new Date(dayEnd.getTime() - 1000),
+      dates: new Set(chunk),
+    });
+  }
+  return out;
+}
 
 export interface DawarichSyncResult {
   /** Days written this run (new or refreshed). */
@@ -137,16 +186,14 @@ export async function syncDawarichRange(
   let visits: Awaited<ReturnType<typeof fetchVisits>> = [];
   let countries: DailyCountries | null = null;
   try {
-    for (let i = 0; i < days.length; i += TIMELINE_CHUNK_DAYS) {
-      const chunk = days.slice(i, i + TIMELINE_CHUNK_DAYS);
-      const { dayStart } = localDayWindow(chunk[0], tz);
-      const { dayEnd } = localDayWindow(chunk[chunk.length - 1], tz);
+    for (const chunk of timelineChunks(days, tz)) {
+      const answered = await fetchTimeline(
+        connection,
+        chunk.startAt,
+        chunk.endAt,
+      );
       timeline = timeline.concat(
-        await fetchTimeline(
-          connection,
-          dayStart,
-          new Date(dayEnd.getTime() - 1000),
-        ),
+        answered.filter((day) => chunk.dates.has(day.date)),
       );
     }
     const rangeStart = localDayWindow(fromDate, tz).dayStart;
diff --git a/src/lib/export/full-backup-payload.ts b/src/lib/export/full-backup-payload.ts
index 5abd26b7d..2879e98f9 100644
--- a/src/lib/export/full-backup-payload.ts
+++ b/src/lib/export/full-backup-payload.ts
@@ -1094,7 +1094,9 @@ export async function buildFullBackupPayload(
             externalId: m.externalId,
             updatedAt: m.updatedAt.toISOString(),
           }
-        : {}),
+        : // No id: a portable restore mints a fresh one. The relation links
+          // still name this drug, so the file says which row they mean.
+          { ref: m.id }),
       name: m.name,
       dose: m.dose,
       active: m.active,
diff --git a/src/lib/export/records-backup.ts b/src/lib/export/records-backup.ts
index f2ca65bd2..239d3e5a3 100644
--- a/src/lib/export/records-backup.ts
+++ b/src/lib/export/records-backup.ts
@@ -83,6 +83,8 @@ function readForExport(
 export interface LabResultBackupEntry {
   /** Present in canonical DR payloads so rows keep a stable identity. */
   id?: string;
+  /** The source row's id in a portable payload, for the relation links. */
+  ref?: string;
   panel: string | null;
   analyte: string;
   value: number | null;
@@ -455,6 +457,9 @@ export async function buildRecordsBackupSection(
           note: null,
         }
       : {
+          // No id: a portable restore mints a fresh one. The relation links
+          // still name this row, so the file says which row they mean.
+          ref: r.id,
           note: readForExport(
             r.noteEncrypted,
             "lab note",
diff --git a/src/lib/export/relation-links-backup.ts b/src/lib/export/relation-links-backup.ts
index e9a0f91c4..ed9ab58ce 100644
--- a/src/lib/export/relation-links-backup.ts
+++ b/src/lib/export/relation-links-backup.ts
@@ -244,6 +244,19 @@ function placeable<T>(
   return kept;
 }
 
+/**
+ * The id each medication and lab result had in the file, to the id the
+ * restore gave it. A portable file carries neither row's id (a restore there
+ * mints fresh ones), only a `ref` naming the source row, so a link that names
+ * the source id is re-pointed through this map. An id the map does not hold is
+ * used as it is: a disaster-recovery file keeps its ids, and a hand-edited one
+ * falls through to the placement check below.
+ */
+export interface RelationLinkSourceIds {
+  medications: ReadonlyMap<string, string>;
+  labResults: ReadonlyMap<string, string>;
+}
+
 /**
  * Re-create the five link tables.
  *
@@ -261,7 +274,13 @@ export async function restoreRelationLinksData(
   ownerId: string,
   payload: RelationLinksBackupSection,
   skips: RestoreSkipLog,
+  sourceIds: RelationLinkSourceIds = {
+    medications: new Map(),
+    labResults: new Map(),
+  },
 ): Promise<RelationLinksRestoreCleared> {
+  const medication = (id: string) => sourceIds.medications.get(id) ?? id;
+  const labResult = (id: string) => sourceIds.labResults.get(id) ?? id;
   const owned = { where: { userId: ownerId } };
   const [c1, c2, c3, c4, c5] = await Promise.all([
     tx.labResultDocumentLink.deleteMany(owned),
@@ -278,7 +297,10 @@ export async function restoreRelationLinksData(
   });
 
   const labDocs = placeable(
-    payload.labResultDocumentLinks,
+    payload.labResultDocumentLinks.map((r) => ({
+      ...r,
+      labResultId: labResult(r.labResultId),
+    })),
     [
       [(r) => r.labResultId, written.labResults],
       [(r) => r.documentId, written.documents],
@@ -297,7 +319,10 @@ export async function restoreRelationLinksData(
   }
 
   const encounterMeds = placeable(
-    payload.encounterMedicationLinks,
+    payload.encounterMedicationLinks.map((r) => ({
+      ...r,
+      medicationId: medication(r.medicationId),
+    })),
     [
       [(r) => r.encounterId, written.encounters],
       [(r) => r.medicationId, written.medications],
@@ -335,7 +360,10 @@ export async function restoreRelationLinksData(
   }
 
   const medDocs = placeable(
-    payload.medicationDocumentLinks,
+    payload.medicationDocumentLinks.map((r) => ({
+      ...r,
+      medicationId: medication(r.medicationId),
+    })),
     [
       [(r) => r.medicationId, written.medications],
       [(r) => r.documentId, written.documents],
@@ -354,7 +382,10 @@ export async function restoreRelationLinksData(
   }
 
   const medConditions = placeable(
-    payload.medicationConditionLinks,
+    payload.medicationConditionLinks.map((r) => ({
+      ...r,
+      medicationId: medication(r.medicationId),
+    })),
     [
       [(r) => r.medicationId, written.medications],
       [(r) => r.episodeId, written.episodes],
diff --git a/src/lib/export/restore-backup.ts b/src/lib/export/restore-backup.ts
index 143689656..c860002cf 100644
--- a/src/lib/export/restore-backup.ts
+++ b/src/lib/export/restore-backup.ts
@@ -1158,6 +1158,12 @@ export async function restoreBackup(
 
         const medByName = new Map<string, string>();
         const restoredMedicationIds = new Set<string>();
+        // The id each medication had in the source account, to the id it got
+        // here. A disaster-recovery file keeps the id, so this is the
+        // identity there; a portable file names the source row only through
+        // `ref`, and the relation links (written last) are re-pointed through
+        // this map instead of being dropped as "end is gone".
+        const medicationIdBySourceId = new Map<string, string>();
         // The efficacy targets cannot ride inside the medication create
         // beside the schedules and the packs: `biomarkerId` is a foreign key
         // into a catalogue this transaction has not restored yet. They are
@@ -1454,6 +1460,10 @@ export async function restoreBackup(
             },
           });
           restoredMedicationIds.add(created.id);
+          const medicationSourceId = m.id ?? m.ref;
+          if (medicationSourceId) {
+            medicationIdBySourceId.set(medicationSourceId, created.id);
+          }
           if (!medByName.has(m.name)) medByName.set(m.name, created.id);
           // v1.39.4 — the clinical category lives in its own table keyed on the
           // medication id, so it is written after the row exists and inside the
@@ -2172,6 +2182,8 @@ export async function restoreBackup(
         // `committedRecordId` names the lab result it was committed to, and
         // that reference is resolved against the rows this loop writes.
         const restoredLabResultIds = new Set<string>();
+        // Same arrangement as `medicationIdBySourceId`, for the lab results.
+        const labResultIdBySourceId = new Map<string, string>();
         // What each drug was supposed to move, written now that both ends
         // exist. The drugs came back several hundred lines up and the
         // analytes only just now, and one of those two had to move for a
@@ -2280,6 +2292,9 @@ export async function restoreBackup(
             },
           });
           restoredLabResultIds.add(createdLab.id);
+          const labSourceId = lab.id ?? lab.ref;
+          if (labSourceId)
+            labResultIdBySourceId.set(labSourceId, createdLab.id);
         }
 
         reportSection("illness");
@@ -2683,6 +2698,10 @@ export async function restoreBackup(
           ownerId,
           payload,
           skips,
+          {
+            medications: medicationIdBySourceId,
+            labResults: labResultIdBySourceId,
+          },
         );
 
         // The Coach's threads, turns and document links. AFTER the
diff --git a/src/lib/insights/status-invalidation.ts b/src/lib/insights/status-invalidation.ts
index 033b289fa..d3f818ccf 100644
--- a/src/lib/insights/status-invalidation.ts
+++ b/src/lib/insights/status-invalidation.ts
@@ -63,7 +63,8 @@ export type StatusIngestSource = IngestSource;
  * day. Since v1.43 (#1217) the gap is chosen by source and by the record's AI
  * usage level (`statusRegenGapHours` in `@/lib/ai/usage-level`): a hand entry
  * keeps the hour (six on `frugal`), a sync waits the operator's sync gap on
- * `standard` (default six hours), a day on `frugal`, the hour on `full`.
+ * `standard` (default six hours), a day on `frugal` (or the operator's gap
+ * when it is longer), the hour on `full`.
  */
 async function ingestRegenGapMs(
   userId: string,
diff --git a/src/lib/jobs/__tests__/pulse-day-mean-rebuild-state.test.ts b/src/lib/jobs/__tests__/pulse-day-mean-rebuild-state.test.ts
new file mode 100644
index 000000000..1873d5653
--- /dev/null
+++ b/src/lib/jobs/__tests__/pulse-day-mean-rebuild-state.test.ts
@@ -0,0 +1,73 @@
+/**
+ * The pulse rollup rebuild is on the record: a stopped one reads as failed
+ * beside the dry run, so the admin card can offer it again.
+ */
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+const db = vi.hoisted(() => {
+  let stored: unknown = {
+    status: "done",
+    startedAt: "2026-10-15T08:00:00.000Z",
+    finishedAt: "2026-10-15T08:10:00.000Z",
+    result: null,
+  };
+  return {
+    get stored() {
+      return stored;
+    },
+    reset(value: unknown) {
+      stored = value;
+    },
+    appSettings: {
+      findUnique: vi.fn(async () => ({ pulseDayMeanDryRunJson: stored })),
+      upsert: vi.fn(
+        async (args: { update: { pulseDayMeanDryRunJson: unknown } }) => {
+          stored = args.update.pulseDayMeanDryRunJson;
+          return {};
+        },
+      ),
+    },
+    $queryRaw: vi.fn(),
+  };
+});
+
+vi.mock("@/lib/db", () => ({
+  prisma: { appSettings: db.appSettings, $queryRaw: db.$queryRaw },
+}));
+vi.mock("@/lib/jobs/job-lock", () => ({
+  withJobLock: async (_key: string, run: () => Promise<unknown>) => ({
+    ran: true,
+    result: await run(),
+  }),
+}));
+vi.mock("@/lib/logging/signal", () => ({ logCaught: vi.fn() }));
+
+import { handlePulseDayMean } from "../pulse-day-mean";
+
+const job = [{ data: { action: "rebuild" } }] as never;
+
+beforeEach(() => {
+  db.$queryRaw.mockReset();
+});
+
+describe("the pulse rebuild state", () => {
+  it("records a stopped rebuild as failed, beside the dry run", async () => {
+    db.$queryRaw.mockRejectedValue(new Error("statement timeout"));
+    const outcome = await handlePulseDayMean(job);
+    expect(outcome.ok).toBe(false);
+    expect(db.stored).toMatchObject({
+      status: "done",
+      rebuild: { status: "failed", accounts: null },
+    });
+  });
+
+  it("records a finished rebuild with what it did", async () => {
+    // No account holds pulse: done, nothing written.
+    db.$queryRaw.mockResolvedValue([]);
+    const outcome = await handlePulseDayMean(job);
+    expect(outcome.ok).toBe(true);
+    expect(db.stored).toMatchObject({
+      rebuild: { status: "done", accounts: 0, rows: 0 },
+    });
+  });
+});
diff --git a/src/lib/jobs/pulse-day-mean.ts b/src/lib/jobs/pulse-day-mean.ts
index f90d4a3a6..c5f9bc758 100644
--- a/src/lib/jobs/pulse-day-mean.ts
+++ b/src/lib/jobs/pulse-day-mean.ts
@@ -41,10 +41,12 @@ import {
 } from "@/lib/rollups/measurement-rollups";
 import {
   PULSE_DAY_MEAN_CHANGE_BUCKETS,
-  pulseDayMeanDryRunSchema,
+  pulseDayMeanStoredRunsSchema,
   type PulseDayMeanDryRun,
   type PulseDayMeanDryRunResult,
+  type PulseDayMeanRebuild,
   type PulseDayMeanStatus,
+  type PulseDayMeanStoredRuns,
 } from "@/lib/validations/pulse-day-mean";
 
 export const PULSE_DAY_MEAN_QUEUE = "pulse-day-mean";
@@ -92,15 +94,26 @@ export async function isPulseDayMeanRunActive(): Promise<boolean> {
   }
 }
 
-function parseDryRun(value: unknown): PulseDayMeanDryRun | null {
-  const parsed = pulseDayMeanDryRunSchema.safeParse(value);
+function parseStoredRuns(value: unknown): PulseDayMeanStoredRuns | null {
+  const parsed = pulseDayMeanStoredRunsSchema.safeParse(value);
   return parsed.success ? parsed.data : null;
 }
 
+/** A run still marked queued or running with no live job behind it failed. */
+function settle<T extends { status: PulseDayMeanDryRun["status"] }>(
+  run: T | null,
+  active: boolean,
+): T | null {
+  if (run && (run.status === "queued" || run.status === "running") && !active) {
+    return { ...run, status: "failed" };
+  }
+  return run;
+}
+
 /**
- * The stored method, when it was applied, and the last dry run. A dry run
- * still marked queued or running with no live job behind it (the worker went
- * away mid-run) reads as failed.
+ * The stored method, when it was applied, the last dry run and the last
+ * rebuild. A run still marked queued or running with no live job behind it
+ * (the worker went away mid-run) reads as failed.
  */
 export async function readPulseDayMeanStatus(): Promise<PulseDayMeanStatus> {
   const row = await prisma.appSettings.findUnique({
@@ -111,32 +124,59 @@ export async function readPulseDayMeanStatus(): Promise<PulseDayMeanStatus> {
       pulseDayMeanDryRunJson: true,
     },
   });
-  let lastDryRun = parseDryRun(row?.pulseDayMeanDryRunJson ?? null);
-  if (
-    lastDryRun &&
-    (lastDryRun.status === "queued" || lastDryRun.status === "running") &&
-    !(await isPulseDayMeanRunActive())
-  ) {
-    lastDryRun = { ...lastDryRun, status: "failed" };
-  }
+  const stored = parseStoredRuns(row?.pulseDayMeanDryRunJson ?? null);
+  const pending =
+    stored !== null &&
+    [stored.status, stored.rebuild?.status].some(
+      (status) => status === "queued" || status === "running",
+    );
+  const active = pending ? await isPulseDayMeanRunActive() : false;
+  const { rebuild, ...dryRun } = stored ?? { rebuild: null };
   return {
     method: isPulseDayMeanMethod(row?.pulseDayMeanMethod)
       ? row.pulseDayMeanMethod
       : DEFAULT_PULSE_DAY_MEAN_METHOD,
     appliedAt: row?.pulseDayMeanAppliedAt?.toISOString() ?? null,
-    lastDryRun,
+    lastDryRun: stored ? settle(dryRun as PulseDayMeanDryRun, active) : null,
+    lastRebuild: settle(rebuild ?? null, active),
   };
 }
 
-/** The only write a dry run makes. */
-export async function writeDryRun(run: PulseDayMeanDryRun): Promise<void> {
+async function readStoredRuns(): Promise<PulseDayMeanStoredRuns | null> {
+  const row = await prisma.appSettings.findUnique({
+    where: { id: "singleton" },
+    select: { pulseDayMeanDryRunJson: true },
+  });
+  return parseStoredRuns(row?.pulseDayMeanDryRunJson ?? null);
+}
+
+async function writeStoredRuns(runs: PulseDayMeanStoredRuns): Promise<void> {
   await prisma.appSettings.upsert({
     where: { id: "singleton" },
-    create: { id: "singleton", pulseDayMeanDryRunJson: run },
-    update: { pulseDayMeanDryRunJson: run },
+    create: { id: "singleton", pulseDayMeanDryRunJson: runs },
+    update: { pulseDayMeanDryRunJson: runs },
   });
 }
 
+/** The only write a dry run makes. The last rebuild stays as it was. */
+export async function writeDryRun(run: PulseDayMeanDryRun): Promise<void> {
+  const stored = await readStoredRuns();
+  await writeStoredRuns({ ...run, rebuild: stored?.rebuild ?? null });
+}
+
+/**
+ * Record the rebuild's state beside the dry run it followed. A rebuild only
+ * runs after an apply, and an apply needs a dry run, so there is always one
+ * to sit beside; without it there is nothing to record against.
+ */
+export async function writeRebuild(
+  rebuild: PulseDayMeanRebuild | null,
+): Promise<void> {
+  const stored = await readStoredRuns();
+  if (!stored) return;
+  await writeStoredRuns({ ...stored, rebuild });
+}
+
 // ─── Accounts ─────────────────────────────────────────────
 
 interface PulseAccount {
@@ -354,6 +394,38 @@ export async function rebuildPulseRollups(): Promise<{
 
 // ─── Job ──────────────────────────────────────────────────
 
+async function runRebuild(): Promise<{ accounts: number; rows: number }> {
+  const startedAt = new Date().toISOString();
+  await writeRebuild({
+    status: "running",
+    startedAt,
+    finishedAt: null,
+    accounts: null,
+    rows: null,
+  });
+  try {
+    const done = await rebuildPulseRollups();
+    await writeRebuild({
+      status: "done",
+      startedAt,
+      finishedAt: new Date().toISOString(),
+      ...done,
+    });
+    return done;
+  } catch (err) {
+    await writeRebuild({
+      status: "failed",
+      startedAt,
+      finishedAt: new Date().toISOString(),
+      accounts: null,
+      rows: null,
+    }).catch((writeErr) =>
+      logCaught("pulse_day_mean.rebuild.status_write_failed", writeErr),
+    );
+    throw err;
+  }
+}
+
 async function runDryRun(): Promise<PulseDayMeanDryRunResult> {
   const startedAt = new Date().toISOString();
   await writeDryRun({
@@ -394,7 +466,7 @@ export async function handlePulseDayMean(
     try {
       const guarded = await withJobLock(PULSE_DAY_MEAN_QUEUE, async () =>
         action === "rebuild"
-          ? { kind: "rebuild" as const, ...(await rebuildPulseRollups()) }
+          ? { kind: "rebuild" as const, ...(await runRebuild()) }
           : { kind: "dry_run" as const, result: await runDryRun() },
       );
       if (!guarded.ran) {
diff --git a/src/lib/measurements/__tests__/consolidation-scan-paged.test.ts b/src/lib/measurements/__tests__/consolidation-scan-paged.test.ts
index 1add7ea76..28661fa62 100644
--- a/src/lib/measurements/__tests__/consolidation-scan-paged.test.ts
+++ b/src/lib/measurements/__tests__/consolidation-scan-paged.test.ts
@@ -1,5 +1,10 @@
 import { describe, expect, it, vi } from "vitest";
 
+const appSettings = vi.hoisted(() => ({
+  findUnique: vi.fn(async () => ({ defaultUserTimezone: "America/New_York" })),
+}));
+vi.mock("@/lib/db", () => ({ prisma: { appSettings } }));
+
 import {
   bucketRowsByDay,
   iterateDayBuckets,
@@ -303,4 +308,29 @@ describe("runConsolidation — streamed walk", () => {
     expect(result.stoppedEarly).toBe(true);
     expect(findMany).toHaveBeenCalledTimes(1);
   });
+
+  it("cuts days in the zone the server publishes for the account", async () => {
+    // A stored zone the server sets aside reads as the instance default, as
+    // it does for `/api/auth/me` and the hour-bucket writer; never as the
+    // drain's own fallback.
+    const findMany = buildPagingFindMany(hourlyRows(1));
+    const base = params(findMany, async () => ({
+      kind: "written" as const,
+      sourceRowsRemoved: 0,
+    }));
+    const zones: string[] = [];
+    await runConsolidation({
+      ...base,
+      prismaClient: {
+        user: {
+          findMany: vi
+            .fn()
+            .mockResolvedValue([{ id: "user-2", timezone: null }]),
+        },
+        measurement: { findMany },
+      } as unknown as PrismaClient,
+      onUserStart: ({ tz }) => zones.push(tz),
+    });
+    expect(zones).toEqual(["America/New_York"]);
+  });
 });
diff --git a/src/lib/measurements/__tests__/intraday-hour-buckets.test.ts b/src/lib/measurements/__tests__/intraday-hour-buckets.test.ts
index e7008f973..53eda2b47 100644
--- a/src/lib/measurements/__tests__/intraday-hour-buckets.test.ts
+++ b/src/lib/measurements/__tests__/intraday-hour-buckets.test.ts
@@ -2,6 +2,7 @@ import { describe, expect, it } from "vitest";
 
 import {
   classifyHourBucket,
+  combineHourWrites,
   cumulativeFromIncrements,
   HOUR_NOT_POSTED,
   isStartOfLocalHour,
@@ -158,3 +159,38 @@ describe("cumulativeFromIncrements", () => {
     expect(postedHours(increments)).toBe(3);
   });
 });
+
+describe("combineHourWrites", () => {
+  const entry = (index: number, startIso: string, value: number) => ({
+    index,
+    type: "ACTIVITY_STEPS" as const,
+    hkIdentifier: "HKQuantityTypeIdentifierStepCount",
+    dateKey: "2026-10-25",
+    hour: 2,
+    value,
+    startMs: Date.parse(startIso),
+  });
+
+  it("sums the two starts of the hour a fall-back day repeats", () => {
+    // Berlin, 2026-10-25: 02:00 CEST is 00:00Z, 02:00 CET is 01:00Z.
+    const { writes, superseded } = combineHourWrites([
+      entry(0, "2026-10-25T00:00:00Z", 300),
+      entry(1, "2026-10-25T01:00:00Z", 120),
+    ]);
+    expect(superseded).toEqual([]);
+    expect(writes).toHaveLength(1);
+    expect(writes[0].write.value).toBe(420);
+    expect(writes[0].contributors.sort()).toEqual([0, 1]);
+  });
+
+  it("lets the last entry of one start win, and marks the earlier one", () => {
+    const { writes, superseded } = combineHourWrites([
+      entry(0, "2026-10-25T00:00:00Z", 300),
+      entry(1, "2026-10-25T00:00:00Z", 310),
+      entry(2, "2026-10-25T01:00:00Z", 120),
+    ]);
+    expect(superseded).toEqual([0]);
+    expect(writes[0].write.value).toBe(430);
+    expect(writes[0].write.index).toBe(2);
+  });
+});
diff --git a/src/lib/measurements/consolidation-base.ts b/src/lib/measurements/consolidation-base.ts
index 0f1b21a73..2b717a0d3 100644
--- a/src/lib/measurements/consolidation-base.ts
+++ b/src/lib/measurements/consolidation-base.ts
@@ -44,6 +44,7 @@ import {
   type PerSampleRow,
 } from "./consolidation-tz";
 import { DEFAULT_TIMEZONE } from "@/lib/tz/format";
+import { resolveStoredTimezone } from "@/lib/tz/resolver";
 
 // Re-export the shared grace constant so the established
 // `consolidation-base` import surface stays unchanged for callers.
@@ -454,7 +455,12 @@ export async function runConsolidation<TType extends MeasurementType>(
   const users = await loadConsolidationUsers(prismaClient, options.userId);
 
   walk: for (const user of users) {
-    const tz = resolveUserTimezone(user.timezone);
+    // The zone the server publishes for the account (`/api/auth/me`), which
+    // is the zone the phone cuts its `stats:` days and hour buckets in and
+    // the one the hour-bucket writer stores them under. A stored value the
+    // server sets aside falls back to the instance default here as it does
+    // there, never to a second fallback of the drain's own.
+    const tz = await resolveStoredTimezone(user.timezone);
     // Aligned to the account's local day: a day is folded whole or not at
     // all, never in two runs (see `foldBoundary`).
     const cutoffAt =
diff --git a/src/lib/measurements/intraday-hour-buckets.ts b/src/lib/measurements/intraday-hour-buckets.ts
index 0c9940b29..2bd5a0830 100644
--- a/src/lib/measurements/intraday-hour-buckets.ts
+++ b/src/lib/measurements/intraday-hour-buckets.ts
@@ -177,6 +177,59 @@ export interface HourBucketWrite {
   value: number;
 }
 
+/** One hour as the batch reads it, with the instant it starts at. */
+export interface HourBucketEntry extends HourBucketWrite {
+  /** `startDate` of the entry, in epoch milliseconds. */
+  startMs: number;
+}
+
+/**
+ * A batch's hours reduced to one write per slot.
+ *
+ * A slot is one local hour of one day. On the autumn day the clock falls
+ * back, one local hour happens twice, and the phone hands over both under
+ * the same key (`…T02`), each with its own `startDate`: they are two hours of
+ * activity and are summed into the slot rather than the second replacing the
+ * first. Entries that repeat the same start are the same hour posted twice,
+ * and the last one counts, as for every other slot.
+ *
+ * `contributors` are the entries whose values the write carries (they take
+ * its status); `superseded` are the earlier repeats of one start.
+ *
+ * The slot holds one number, so the sum is only as complete as the batch
+ * that carries it: both occurrences have to arrive together. The phone posts
+ * a day's hours in one batch, and a later batch that carries only one of
+ * them replaces the slot with that one, as a re-post of any slot does.
+ */
+export function combineHourWrites(entries: readonly HourBucketEntry[]): {
+  writes: Array<{ write: HourBucketWrite; contributors: number[] }>;
+  superseded: number[];
+} {
+  const slots = new Map<string, Map<number, HourBucketEntry>>();
+  const superseded: number[] = [];
+  for (const entry of entries) {
+    const key = `${entry.type}::${entry.dateKey}::${entry.hour}`;
+    const byStart = slots.get(key) ?? new Map<number, HourBucketEntry>();
+    const earlier = byStart.get(entry.startMs);
+    if (earlier) superseded.push(earlier.index);
+    byStart.set(entry.startMs, entry);
+    slots.set(key, byStart);
+  }
+  const writes = [...slots.values()].map((byStart) => {
+    const occurrences = [...byStart.values()];
+    const last = occurrences.reduce((a, b) => (b.index > a.index ? b : a));
+    const { startMs: _start, ...base } = last;
+    return {
+      write: {
+        ...base,
+        value: occurrences.reduce((sum, entry) => sum + entry.value, 0),
+      },
+      contributors: occurrences.map((entry) => entry.index),
+    };
+  });
+  return { writes, superseded };
+}
+
 /**
  * Store a batch's hours, one transaction per (type, day). Returns the
  * per-entry status: `inserted` for the first value of a slot, `updated` when
diff --git a/src/lib/medications/ingredients/__tests__/recompute-on-switch.test.ts b/src/lib/medications/ingredients/__tests__/recompute-on-switch.test.ts
new file mode 100644
index 000000000..312073ad2
--- /dev/null
+++ b/src/lib/medications/ingredients/__tests__/recompute-on-switch.test.ts
@@ -0,0 +1,41 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+const send = vi.hoisted(() => vi.fn(async (..._args: unknown[]) => "job-1"));
+vi.mock("@/lib/jobs/boss-instance", () => ({
+  getGlobalBoss: () => ({ send }),
+}));
+vi.mock("@/lib/db", () => ({ prisma: {} }));
+vi.mock("@/lib/logging/context", () => ({ annotate: vi.fn() }));
+vi.mock("../recompute", () => ({
+  recomputeSupplementNutrientsForUser: vi.fn(),
+}));
+
+import { recomputeSupplementsIfNutrientsTurnedOn } from "../recompute-job";
+
+beforeEach(() => send.mockClear());
+
+describe("the supplement recompute after a module switch", () => {
+  it("queues the full recompute when nutrients comes back on", async () => {
+    await recomputeSupplementsIfNutrientsTurnedOn(
+      "user-1",
+      { nutrients: false },
+      { nutrients: true },
+    );
+    expect(send).toHaveBeenCalledTimes(1);
+    expect(send.mock.calls[0]?.[1]).toMatchObject({ userId: "user-1" });
+  });
+
+  it("queues nothing when the switch did not move to on", async () => {
+    await recomputeSupplementsIfNutrientsTurnedOn(
+      "user-1",
+      { nutrients: true },
+      { nutrients: true, mood: false },
+    );
+    await recomputeSupplementsIfNutrientsTurnedOn(
+      "user-1",
+      { nutrients: true },
+      { nutrients: false },
+    );
+    expect(send).not.toHaveBeenCalled();
+  });
+});
diff --git a/src/lib/medications/ingredients/recompute-job.ts b/src/lib/medications/ingredients/recompute-job.ts
index 2f0e00759..e42624420 100644
--- a/src/lib/medications/ingredients/recompute-job.ts
+++ b/src/lib/medications/ingredients/recompute-job.ts
@@ -17,6 +17,8 @@
 import { prisma } from "@/lib/db";
 import { getGlobalBoss } from "@/lib/jobs/boss-instance";
 import { annotate } from "@/lib/logging/context";
+import { normalisePrefs } from "@/lib/modules/gate";
+import { isOptInModule } from "@/lib/modules/registry";
 
 import { recomputeSupplementNutrientsForUser } from "./recompute";
 
@@ -67,6 +69,28 @@ export async function enqueueSupplementNutrientRecompute(
   return { queued: false };
 }
 
+/** The account's own `nutrients` switch in a stored preference blob. */
+function nutrientsSwitchOn(prefs: unknown): boolean {
+  const value = normalisePrefs(prefs).nutrients;
+  return isOptInModule("nutrients") ? value === true : value !== false;
+}
+
+/**
+ * Queue the full recompute when a preference write turned the account's
+ * `nutrients` switch back on. While the module is off nothing is written
+ * (`recompute.ts`), so a dose removed, a composition changed or a zone moved
+ * in the meantime left the stored rows as they were; the first pass after
+ * the switch puts them right instead of the next unrelated change.
+ */
+export async function recomputeSupplementsIfNutrientsTurnedOn(
+  userId: string,
+  before: unknown,
+  after: unknown,
+): Promise<void> {
+  if (nutrientsSwitchOn(before) || !nutrientsSwitchOn(after)) return;
+  await enqueueSupplementNutrientRecompute(userId);
+}
+
 /** The pg-boss handler body for one job. */
 export async function runSupplementNutrientRecompute(
   payload: SupplementNutrientRecomputePayload,
diff --git a/src/lib/nightscout/__tests__/client.test.ts b/src/lib/nightscout/__tests__/client.test.ts
index 7005a56e3..1a08b9004 100644
--- a/src/lib/nightscout/__tests__/client.test.ts
+++ b/src/lib/nightscout/__tests__/client.test.ts
@@ -18,11 +18,10 @@ import {
 } from "../client";
 
 function jsonResponse(status: number, body: unknown): Response {
-  return {
+  return new Response(JSON.stringify(body), {
     status,
-    ok: status >= 200 && status < 300,
-    json: async () => body,
-  } as unknown as Response;
+    headers: { "content-type": "application/json" },
+  });
 }
 
 const SAMPLE: NightscoutSgvEntry[] = [
@@ -295,4 +294,27 @@ describe("fetchSgvEntries — SSRF gating", () => {
     expect(entries).toHaveLength(2);
     expect(entries[0]!.sgv).toBe(112);
   });
+
+  it("stops reading an answer that runs past the cap", async () => {
+    // Streamed without a length, so only the running count can catch it.
+    const chunk = new Uint8Array(1024 * 1024).fill(32);
+    let sent = 0;
+    const body = new ReadableStream<Uint8Array>({
+      pull(controller) {
+        sent += 1;
+        if (sent > 64) controller.close();
+        else controller.enqueue(chunk);
+      },
+    });
+    safeFetchMock.mockResolvedValue(new Response(body, { status: 200 }));
+    await expect(
+      fetchSgvEntries({
+        baseUrl: "https://ns.example.com",
+        token: "",
+        count: 10,
+        allowPrivateHost: false,
+      }),
+    ).rejects.toThrow(/oversized/);
+    expect(sent).toBeLessThan(64);
+  });
 });
diff --git a/src/lib/nightscout/client.ts b/src/lib/nightscout/client.ts
index e41be4bcd..acaf4454b 100644
--- a/src/lib/nightscout/client.ts
+++ b/src/lib/nightscout/client.ts
@@ -23,6 +23,7 @@
 import { createHash } from "node:crypto";
 
 import type { MeasurementType } from "@/generated/prisma/client";
+import { readBodyCapped } from "@/lib/http/read-capped";
 import { safeFetch } from "@/lib/safe-fetch";
 import {
   configuredNightscoutPrivateOrigins,
@@ -33,6 +34,7 @@ import { SafeFetchError } from "@/lib/safe-fetch";
 
 /** Default per-request timeout for a Nightscout fetch. */
 const NIGHTSCOUT_TIMEOUT_MS = 15_000;
+const NIGHTSCOUT_MAX_BODY_BYTES = 8 * 1024 * 1024;
 
 /** Canonical glucose storage unit. Nightscout SGV is always mg/dL on the wire. */
 export const NIGHTSCOUT_GLUCOSE_UNIT = "mg/dL" as const;
@@ -259,9 +261,15 @@ export async function fetchSgvEntries(
     );
   }
 
+  // Bounded while reading: `count` entries are a few hundred bytes each, so
+  // the cap is far above any honest answer and stops one that never ends.
+  const read = await readBodyCapped(res, NIGHTSCOUT_MAX_BODY_BYTES);
+  if (!read.ok) {
+    throw new NightscoutApiError("Nightscout returned an oversized body", null);
+  }
   let body: unknown;
   try {
-    body = await res.json();
+    body = JSON.parse(read.text) as unknown;
   } catch {
     throw new NightscoutApiError("Nightscout returned a non-JSON body", null);
   }
diff --git a/src/lib/openapi/routes/ai-usage.ts b/src/lib/openapi/routes/ai-usage.ts
index 1711b4267..e70982236 100644
--- a/src/lib/openapi/routes/ai-usage.ts
+++ b/src/lib/openapi/routes/ai-usage.ts
@@ -16,7 +16,7 @@ import { dataEnvelope, errorEnvelope, stdResponses } from "./shared";
 export const aiUsageLevel = z.enum(AI_USAGE_LEVELS).meta({
   id: "AiUsageLevel",
   description:
-    "How much background AI work an account asks for, in ceiling order. `frugal`: status cards regenerate at most every 24 h after a sync and every 6 h after a hand entry, a smaller snapshot, no background reasoning, a short daily briefing (`recommendations` holds at most one entry, every field stays present), no automatic document reading, no supplement label reading. `standard`: the operator's sync gap (default 6 h), 1 h after a hand entry, background reasoning at the job default. `full`: 1 h after any ingest (the behaviour before v1.43). The interactive Coach is the same at every level.",
+    "How much background AI work an account asks for, in ceiling order. `frugal`: status cards regenerate at most every 24 h after a sync (or the operator's sync gap when that is longer, so never more often than `standard`) and every 6 h after a hand entry, a smaller snapshot, no background reasoning, a short daily briefing (`recommendations` holds at most one entry, every field stays present), no automatic document reading, no supplement label reading. `standard`: the operator's sync gap (default 6 h), 1 h after a hand entry, background reasoning at the job default. `full`: 1 h after any ingest (the behaviour before v1.43). The interactive Coach is the same at every level.",
 });
 
 export const aiUsageState = z
diff --git a/src/lib/openapi/routes/life-circumstances.ts b/src/lib/openapi/routes/life-circumstances.ts
index 590dcabd3..4e92b11e6 100644
--- a/src/lib/openapi/routes/life-circumstances.ts
+++ b/src/lib/openapi/routes/life-circumstances.ts
@@ -119,7 +119,11 @@ const lifeCircumstances = z
       screeningHints: z.array(screeningHint),
     }),
     auditC: z.object({
-      latest: auditCResult.nullable(),
+      latest: auditCResult
+        .nullable()
+        .describe(
+          "The newest entry whose answers this host can open; an entry it cannot open is listed in `history` only.",
+        ),
       history: z
         .array(
           z.object({
@@ -129,7 +133,11 @@ const lifeCircumstances = z
           }),
         )
         .describe("Newest first, at most 20."),
-      offerDue: z.boolean(),
+      offerDue: z
+        .boolean()
+        .describe(
+          "No readable entry, or the newest one whose answers open (the one in `latest`) is 365 days old or older.",
+        ),
       instrument: z.object({
         items: z
           .array(
@@ -268,7 +276,7 @@ export const lifeCircumstancesPaths: NonNullable<ZodOpenApiObject["paths"]> = {
       tags: ["Anamnesis"],
       summary: "Record an AUDIT-C (v1.43)",
       description:
-        "Scores the three items on the server. A replay with the same `externalId` answers 200 with the stored result. 422 `anamnesis.auditC.invalid`; 429 `anamnesis.auditC.rateLimited` after 30 in an hour.",
+        "Scores the three items on the server. The entry is stored in the language of the wording the read route offers this account (`auditC.instrument.locale`), and the cut-off follows that wording; `locale`, when sent, must name that same wording. A replay with the same `externalId` answers 200 with the stored result. 422 `anamnesis.auditC.invalid` (also for another wording than the one offered, and for a `takenAt` more than five minutes ahead of the server clock); 429 `anamnesis.auditC.rateLimited` after 30 in an hour.",
       parameters: [idempotencyKeyParameter],
       requestBody: {
         required: true,
diff --git a/src/lib/profile/audit-c-entries.ts b/src/lib/profile/audit-c-entries.ts
new file mode 100644
index 000000000..ce2d50862
--- /dev/null
+++ b/src/lib/profile/audit-c-entries.ts
@@ -0,0 +1,93 @@
+/**
+ * v1.43 (#614) — which AUDIT-C entry is "the newest" for every reader.
+ *
+ * The answers are sealed; the sum is stored in the clear. An entry whose
+ * answers this host can no longer open (a lost or retired key) still has a
+ * readable sum, so a reader that skipped decryption used to take it as the
+ * newest result while the read route, which needs the answers, showed an
+ * older one beside it and offered the questionnaire again on a date neither
+ * of them showed. One rule now holds everywhere a result is read (the read
+ * route's `latest` and `offerDue`, the Coach block, the doctor report and
+ * FHIR): the newest entry whose answers open. The history keeps listing every
+ * entry by its sum, so an unreadable one can still be found and removed.
+ */
+import { decryptFromBytes } from "@/lib/ai/coach/bytes-codec";
+import { prisma } from "@/lib/db";
+import { getEvent } from "@/lib/logging/context";
+
+/** How many entries back a reader looks for one it can open. */
+export const AUDIT_C_HISTORY_LIMIT = 20;
+
+/**
+ * The sealed answers of one AUDIT-C entry. `schema` versions the blob the same
+ * way the screener responses are versioned.
+ */
+export interface AuditCResponses {
+  schema: 1;
+  items: [number, number, number];
+}
+
+/** Open the sealed AUDIT-C answers; null when this host cannot read them. */
+export function openAuditCResponses(
+  payload: Uint8Array,
+): AuditCResponses | null {
+  try {
+    const parsed = JSON.parse(
+      decryptFromBytes(payload),
+    ) as Partial<AuditCResponses>;
+    const items = parsed.items;
+    if (
+      Array.isArray(items) &&
+      items.length === 3 &&
+      items.every((item) => Number.isInteger(item) && item >= 0 && item <= 4)
+    ) {
+      return { schema: 1, items: items as [number, number, number] };
+    }
+    getEvent()?.addWarning("audit-c responses shape drifted");
+  } catch (error) {
+    getEvent()?.addWarning(
+      `audit-c responses decrypt failed: ${
+        error instanceof Error ? error.message : String(error)
+      }`,
+    );
+  }
+  return null;
+}
+
+/** The first entry of `rows` (newest first) whose answers open. */
+export function newestReadableAuditC<
+  T extends { responsesEncrypted: Uint8Array },
+>(rows: readonly T[]): { row: T; items: [number, number, number] } | null {
+  for (const row of rows) {
+    const opened = openAuditCResponses(row.responsesEncrypted);
+    if (opened) return { row, items: opened.items };
+  }
+  return null;
+}
+
+/** The newest readable entry of a record, for the readers that need one. */
+export async function loadNewestReadableAuditC(userId: string): Promise<{
+  id: string;
+  takenAt: Date;
+  totalScore: number;
+  locale: string;
+  tz: string | null;
+} | null> {
+  const rows = await prisma.mentalHealthAssessment.findMany({
+    where: { userId, instrument: "AUDIT_C", deletedAt: null },
+    orderBy: [{ takenAt: "desc" }, { id: "desc" }],
+    take: AUDIT_C_HISTORY_LIMIT,
+    select: {
+      id: true,
+      takenAt: true,
+      totalScore: true,
+      locale: true,
+      tz: true,
+      responsesEncrypted: true,
+    },
+  });
+  const newest = newestReadableAuditC(rows);
+  if (!newest) return null;
+  const { responsesEncrypted: _sealed, ...entry } = newest.row;
+  return entry;
+}
diff --git a/src/lib/profile/life-circumstances.ts b/src/lib/profile/life-circumstances.ts
index 932fc5455..f9c02127d 100644
--- a/src/lib/profile/life-circumstances.ts
+++ b/src/lib/profile/life-circumstances.ts
@@ -13,8 +13,15 @@
  * Server-only.
  */
 import { Prisma } from "@/generated/prisma/client";
-import { decryptFromBytes, encryptToBytes } from "@/lib/ai/coach/bytes-codec";
+import { encryptToBytes } from "@/lib/ai/coach/bytes-codec";
 import { deriveAgeYears } from "@/lib/ai/coach/about-me";
+import {
+  AUDIT_C_HISTORY_LIMIT,
+  loadNewestReadableAuditC,
+  newestReadableAuditC,
+  openAuditCResponses,
+  type AuditCResponses,
+} from "@/lib/profile/audit-c-entries";
 import { prisma } from "@/lib/db";
 import { getEvent } from "@/lib/logging/context";
 import {
@@ -40,6 +47,7 @@ import {
   computePackYears,
   DEFAULT_FREE_WEEKDAYS,
   isReviewDue,
+  LIFE_KIND_AI_SECTION,
   screeningHints,
   type FreeWeekdaysSpan,
   type PackYears,
@@ -63,22 +71,11 @@ import {
 } from "@/lib/validations/life-circumstances";
 
 const MS_PER_DAY = 86_400_000;
-/** History entries the read route lists, newest first. */
-const AUDIT_C_HISTORY_LIMIT = 20;
 /** AUDIT-C is offered again once the newest entry is this old. */
 const AUDIT_C_OFFER_AFTER_DAYS = 365;
 /** A step count this recent means activity is measured, not only reported. */
 const MEASURED_STEPS_DAYS = 30;
 
-/**
- * The sealed answers of one AUDIT-C entry. `schema` versions the blob the same
- * way the screener responses are versioned.
- */
-interface AuditCResponses {
-  schema: 1;
-  items: [number, number, number];
-}
-
 export interface AuditCResult {
   id: string;
   takenAt: string;
@@ -160,31 +157,6 @@ function latest(...dates: Array<Date | null | undefined>): Date | null {
   return best;
 }
 
-/** Open the sealed AUDIT-C answers; null when this host cannot read them. */
-function openAuditCResponses(payload: Uint8Array): AuditCResponses | null {
-  try {
-    const parsed = JSON.parse(
-      decryptFromBytes(payload),
-    ) as Partial<AuditCResponses>;
-    const items = parsed.items;
-    if (
-      Array.isArray(items) &&
-      items.length === 3 &&
-      items.every((item) => Number.isInteger(item) && item >= 0 && item <= 4)
-    ) {
-      return { schema: 1, items: items as [number, number, number] };
-    }
-    getEvent()?.addWarning("audit-c responses shape drifted");
-  } catch (error) {
-    getEvent()?.addWarning(
-      `audit-c responses decrypt failed: ${
-        error instanceof Error ? error.message : String(error)
-      }`,
-    );
-  }
-  return null;
-}
-
 /** One stored entry scored for a profile sex; the source line follows `labelLocale`. */
 export function toAuditCResult(
   row: { id: string; takenAt: Date; totalScore: number; locale: string },
@@ -243,6 +215,42 @@ export async function loadFreeWeekdaySpans(
   }
 }
 
+/**
+ * The free-weekday answers for a prompt: only when the section they belong
+ * to (`WORK`) is released to the AI. Otherwise the answers are never read or
+ * decrypted for a prompt, and the night-by-night day type falls back to the
+ * weekend default, exactly as for a record that never answered. The data
+ * surfaces (the sleep page, the dashboard) read `loadFreeWeekdaySpans`
+ * directly: they show the person their own answer, no model sees it.
+ */
+export async function loadFreeWeekdaySpansForPrompt(
+  userId: string,
+): Promise<FreeWeekdaysSpan[]> {
+  // Fail-soft and fail-closed, like the read below: a failed switch read
+  // gives the weekend default and never the answers.
+  try {
+    const profile = await prisma.userHealthProfile.findUnique({
+      where: { userId },
+      select: { aiIncludedSections: true },
+    });
+    if (
+      !storedSections(profile?.aiIncludedSections).has(
+        LIFE_KIND_AI_SECTION.FREE_WEEKDAYS,
+      )
+    ) {
+      return [];
+    }
+  } catch (error) {
+    getEvent()?.addWarning(
+      `free weekdays switch read failed: ${
+        error instanceof Error ? error.message : String(error)
+      }`,
+    );
+    return [];
+  }
+  return loadFreeWeekdaySpans(userId);
+}
+
 /** Assemble the whole read model for one record. */
 export async function buildLifeCircumstances(
   userId: string,
@@ -327,17 +335,16 @@ export async function buildLifeCircumstances(
   const factOf = <T>(kind: string): T | null =>
     (facts.find((fact) => fact.kind === kind)?.value as T | undefined) ?? null;
 
-  // AUDIT-C: the newest entry whose answers this host can open is the one
-  // shown in full; the history lists totals, which need no decryption.
-  let latestAuditC: AuditCResult | null = null;
-  for (const row of auditRows) {
-    const opened = openAuditCResponses(row.responsesEncrypted);
-    if (opened) {
-      latestAuditC = toAuditCResult(row, opened.items, sex, locale);
-      break;
-    }
-  }
+  // AUDIT-C: the newest entry whose answers this host can open is the
+  // result, and the offer is timed from it, as every other reader times it
+  // (`audit-c-entries.ts`). The history lists totals, which need no
+  // decryption, so an unreadable entry can still be removed.
+  const readable = newestReadableAuditC(auditRows);
+  const latestAuditC: AuditCResult | null = readable
+    ? toAuditCResult(readable.row, readable.items, sex, locale)
+    : null;
   const newestAuditC = auditRows[0]?.takenAt ?? null;
+  const newestReadableAt = readable?.row.takenAt ?? null;
 
   const sections = LIFE_SECTIONS.map((section) => {
     const sectionFacts = facts.filter((fact) =>
@@ -398,8 +405,8 @@ export async function buildLifeCircumstances(
         totalScore: row.totalScore,
       })),
       offerDue:
-        newestAuditC === null ||
-        now.getTime() - newestAuditC.getTime() >=
+        newestReadableAt === null ||
+        now.getTime() - newestReadableAt.getTime() >=
           AUDIT_C_OFFER_AFTER_DAYS * MS_PER_DAY,
       instrument: auditCInstrument(locale),
     },
@@ -438,6 +445,25 @@ function lifeSwitches(
   ) as Record<LifeAiSection, boolean>;
 }
 
+/**
+ * Serialise every read-modify-write of a profile's AI switches. Two routes
+ * write the one array: this module's switch route and the about-me PUT, which
+ * replaces its eight sections and carries the life switches over as stored.
+ * Without a lock, a PUT that read the array before a concurrent switch-off
+ * committed wrote the old value back and turned the section on again. A
+ * transaction-scoped advisory lock covers the first write too, when there is
+ * no row yet to lock.
+ */
+export async function lockAiSections(
+  tx: Prisma.TransactionClient,
+  userId: string,
+): Promise<void> {
+  await tx.$queryRaw`
+    SELECT 1 AS locked
+    FROM pg_advisory_xact_lock(hashtextextended(${`ai-sections:${userId}`}, 0))
+  `;
+}
+
 /**
  * Switch life-circumstance sections in or out of the AI inclusion list. Only
  * the named sections change; the other stored values, the five about-me
@@ -452,6 +478,7 @@ export async function updateLifeAiSections(
   base?: Date,
 ): Promise<LifeAiSectionsOutcome> {
   return prisma.$transaction(async (tx) => {
+    await lockAiSections(tx, userId);
     const row = await tx.userHealthProfile.findUnique({
       where: { userId },
       select: { aiIncludedSections: true, updatedAt: true },
@@ -646,11 +673,7 @@ export async function loadLifeClinicalSummary(
       },
       select: { kind: true, valueEncrypted: true },
     }),
-    prisma.mentalHealthAssessment.findFirst({
-      where: { userId, instrument: "AUDIT_C", deletedAt: null },
-      orderBy: [{ takenAt: "desc" }, { id: "desc" }],
-      select: { id: true, takenAt: true, totalScore: true, locale: true },
-    }),
+    loadNewestReadableAuditC(userId),
   ]);
   const value = <T>(kind: string): T | null => {
     const row = rows.find((candidate) => candidate.kind === kind);
diff --git a/src/lib/record-settings/modules.ts b/src/lib/record-settings/modules.ts
index 13428f2ad..0c2df12e9 100644
--- a/src/lib/record-settings/modules.ts
+++ b/src/lib/record-settings/modules.ts
@@ -35,6 +35,7 @@
 import { invalidateUserHealthScore } from "@/lib/cache/invalidate";
 import { getOrCreateCycleProfile } from "@/lib/cycle/profile";
 import { prisma, toJson } from "@/lib/db";
+import { recomputeSupplementsIfNutrientsTurnedOn } from "@/lib/medications/ingredients/recompute-job";
 
 export interface RecordModuleWrite {
   /** The record whose modules these are — never the actor, under a switch. */
@@ -59,10 +60,19 @@ export async function writeRecordModulePreferences(
   input: RecordModuleWrite,
 ): Promise<{ cycleTrackingEnabled: boolean | null }> {
   if (input.modulePreferences !== undefined) {
+    const before = await prisma.user.findUnique({
+      where: { id: input.recordId },
+      select: { modulePreferencesJson: true },
+    });
     await prisma.user.update({
       where: { id: input.recordId },
       data: { modulePreferencesJson: toJson(input.modulePreferences) },
     });
+    await recomputeSupplementsIfNutrientsTurnedOn(
+      input.recordId,
+      before?.modulePreferencesJson,
+      input.modulePreferences,
+    );
   }
   if (input.cycleTrackingEnabled !== undefined) {
     await getOrCreateCycleProfile(input.recordId);
diff --git a/src/lib/validations/backup.ts b/src/lib/validations/backup.ts
index bb6ad4f24..46f816c19 100644
--- a/src/lib/validations/backup.ts
+++ b/src/lib/validations/backup.ts
@@ -392,6 +392,10 @@ const customMedicationCategorySchema = z
 const medicationSchema = z
   .object({
     id: z.string().min(1).optional(),
+    // The source row's id in a portable file, which carries no `id`: the
+    // relation links name the medication by it, and the restore re-points
+    // them at the row it writes (`relation-links-backup.ts`).
+    ref: z.string().min(1).optional(),
     name: z.string().min(1),
     dose: z.string(),
     treatmentClass: z.enum(MedicationCategory).optional(),
@@ -1126,6 +1130,8 @@ const appSettingsBackupSchema = z
 const labResultBackupSchema = z
   .object({
     id: z.string().min(1).optional(),
+    // The source row's id in a portable file, as on a medication.
+    ref: z.string().min(1).optional(),
     panel: z.string().nullable().optional(),
     analyte: z.string().min(1),
     value: z.number().nullable().optional(),
diff --git a/src/lib/validations/pulse-day-mean.ts b/src/lib/validations/pulse-day-mean.ts
index f83a65269..5a0c4127b 100644
--- a/src/lib/validations/pulse-day-mean.ts
+++ b/src/lib/validations/pulse-day-mean.ts
@@ -9,7 +9,9 @@ import { z } from "zod/v4";
 import type { PulseDayMeanMethod } from "@/lib/measurements/day-statistic";
 
 export const pulseDayMeanActionSchema = z.strictObject({
-  action: z.enum(["dry_run", "apply"]),
+  // `rebuild` runs the recalculation again after an apply whose rebuild
+  // stopped, without a new dry run: the method is already switched.
+  action: z.enum(["dry_run", "apply", "rebuild"]),
 });
 export type PulseDayMeanActionInput = z.infer<typeof pulseDayMeanActionSchema>;
 
@@ -46,20 +48,47 @@ const dryRunResultSchema = z.object({
 });
 export type PulseDayMeanDryRunResult = z.infer<typeof dryRunResultSchema>;
 
-/** `app_settings.pulse_day_mean_dry_run_json`: the last dry run. */
-export const pulseDayMeanDryRunSchema = z.object({
-  status: z.enum(["queued", "running", "done", "failed"]),
+const runStatusSchema = z.enum(["queued", "running", "done", "failed"]);
+
+/** The last dry run. */
+const pulseDayMeanDryRunSchema = z.object({
+  status: runStatusSchema,
   startedAt: z.string(),
   finishedAt: z.string().nullable(),
   result: dryRunResultSchema.nullable(),
 });
 export type PulseDayMeanDryRun = z.infer<typeof pulseDayMeanDryRunSchema>;
 
+/** The last recalculation of the pulse rollups after an apply. */
+const pulseDayMeanRebuildSchema = z.object({
+  status: runStatusSchema,
+  startedAt: z.string(),
+  finishedAt: z.string().nullable(),
+  /** Accounts and rollup rows done; null until the run has finished. */
+  accounts: z.number().int().nonnegative().nullable(),
+  rows: z.number().int().nonnegative().nullable(),
+});
+export type PulseDayMeanRebuild = z.infer<typeof pulseDayMeanRebuildSchema>;
+
+/**
+ * `app_settings.pulse_day_mean_dry_run_json`: the last dry run, and beside it
+ * the last rebuild. One column holds both, because the rebuild only ever
+ * follows an apply that needed a dry run; a blob written before the rebuild
+ * was recorded has no `rebuild` key and reads as none.
+ */
+export const pulseDayMeanStoredRunsSchema = pulseDayMeanDryRunSchema.extend({
+  rebuild: pulseDayMeanRebuildSchema.nullable().optional(),
+});
+export type PulseDayMeanStoredRuns = z.infer<
+  typeof pulseDayMeanStoredRunsSchema
+>;
+
 /** `GET /api/admin/maintenance/pulse-day-mean`. */
 export interface PulseDayMeanStatus {
   method: PulseDayMeanMethod;
   appliedAt: string | null;
   lastDryRun: PulseDayMeanDryRun | null;
+  lastRebuild: PulseDayMeanRebuild | null;
 }
 
 /** How old a finished dry run may be for the apply step to accept it. */
diff --git a/tests/integration/dawarich-context.test.ts b/tests/integration/dawarich-context.test.ts
index f91aba072..04a1e1855 100644
--- a/tests/integration/dawarich-context.test.ts
+++ b/tests/integration/dawarich-context.test.ts
@@ -134,6 +134,10 @@ function residency(year: number) {
   };
 }
 
+/** A timeline that answers from the window, set by the test that needs it. */
+let timelineAnswer: ((url: URL) => { status: number; body: unknown }) | null =
+  null;
+
 function answer(req: IncomingMessage): { status: number; body: unknown } {
   const url = new URL(req.url ?? "/", "http://stub");
   requests.push(url.pathname);
@@ -144,6 +148,7 @@ function answer(req: IncomingMessage): { status: number; body: unknown } {
     case "/api/v1/users/me":
       return { status: 200, body: { email: "someone@example.test" } };
     case "/api/v1/timeline":
+      if (timelineAnswer) return timelineAnswer(url);
       return { status: 200, body: { days: TIMELINE_DAYS } };
     case "/api/v1/visits":
       return { status: 200, body: VISITS };
@@ -188,6 +193,7 @@ beforeEach(async () => {
   await truncateAllTables(prisma);
   cookieJar.clear();
   requests.length = 0;
+  timelineAnswer = null;
   process.env.DAWARICH_PRIVATE_ORIGINS = origin;
   await prisma.user.create({
     data: {
@@ -516,3 +522,73 @@ describe("Disconnect", () => {
     expect(await storedDays()).toHaveLength(0);
   });
 });
+
+describe("Timeline days across a request boundary", () => {
+  /**
+   * The instance as Dawarich builds a timeline (`Timeline::DayAssembler`):
+   * days in the Dawarich account's own zone, here New York while the
+   * HealthLog account is in Berlin, and only what lies inside the requested
+   * window is summed, so a window cut at Berlin midnight answers the New
+   * York day it cuts with a fragment. Every New York day holds a 1 km trip
+   * at 02:00 and a 10 km trip at 20:00 (02:00 the next morning in Berlin),
+   * so a complete day is 11 km and a fragment 1 or 10. A window longer than
+   * 31 days is refused as the instance refuses it.
+   */
+  const INSTANCE_TZ = "America/New_York";
+  const nyDate = (ms: number) =>
+    new Date(ms).toLocaleDateString("en-CA", { timeZone: INSTANCE_TZ });
+  const nyInstant = (date: string, hour: number) => {
+    // New York is UTC-4 for the whole summer range this test reads.
+    return Date.parse(`${date}T${String(hour).padStart(2, "0")}:00:00-04:00`);
+  };
+  function instanceTimeline(url: URL) {
+    const startMs = Date.parse(url.searchParams.get("start_at") ?? "");
+    const endMs = Date.parse(url.searchParams.get("end_at") ?? "");
+    if (endMs - startMs > 31 * 24 * 3_600_000) {
+      return { status: 400, body: { error: "range too large" } };
+    }
+    const totals = new Map<string, number>();
+    const from = new Date(startMs - 2 * 86_400_000);
+    for (let d = from.getTime(); d <= endMs + 86_400_000; d += 86_400_000) {
+      const date = nyDate(d);
+      for (const [hour, km] of [
+        [2, 1],
+        [20, 10],
+      ] as const) {
+        const at = nyInstant(date, hour);
+        if (at < startMs || at > endMs) continue;
+        totals.set(date, (totals.get(date) ?? 0) + km);
+      }
+    }
+    return {
+      status: 200,
+      body: {
+        days: [...totals].map(([date, km]) => ({
+          date,
+          summary: {
+            total_distance: km,
+            places_visited: 0,
+            mode_distances: {},
+          },
+        })),
+      },
+    };
+  }
+
+  it("never lets a boundary fragment overwrite a complete day", async () => {
+    timelineAnswer = instanceTimeline;
+    expect((await connect()).status).toBe(200);
+    await getPrismaClient().dawarichDay.deleteMany({});
+
+    const { syncDawarichRange } = await import("@/lib/dawarich/sync");
+    // Sixty days: three timeline requests, two boundaries between them.
+    await syncDawarichRange(USER_ID, "2026-06-01", "2026-07-30");
+
+    const days = await storedDays();
+    expect(days).toHaveLength(60);
+    const wrong = days
+      .filter((d) => d.totalDistanceM !== 11_000)
+      .map((d) => `${d.date}=${d.totalDistanceM}`);
+    expect(wrong).toEqual([]);
+  });
+});
diff --git a/tests/integration/life-circumstances.test.ts b/tests/integration/life-circumstances.test.ts
index fa5088705..508c8a53c 100644
--- a/tests/integration/life-circumstances.test.ts
+++ b/tests/integration/life-circumstances.test.ts
@@ -297,6 +297,34 @@ describe("life-circumstance facts", () => {
   });
 });
 
+describe("free weekdays in a prompt", () => {
+  it("are read for a prompt only while the work section is released", async () => {
+    const user = await makeUser();
+    await signIn(user.id);
+    await data(
+      await postFact({ kind: "FREE_WEEKDAYS", value: VALUES.FREE_WEEKDAYS }),
+      201,
+    );
+    const { loadFreeWeekdaySpans, loadFreeWeekdaySpansForPrompt } =
+      await import("@/lib/profile/life-circumstances");
+
+    await getPrismaClient().userHealthProfile.upsert({
+      where: { userId: user.id },
+      create: { userId: user.id, aiIncludedSections: ["WORK"] },
+      update: { aiIncludedSections: ["WORK"] },
+    });
+    expect(await loadFreeWeekdaySpansForPrompt(user.id)).toHaveLength(1);
+
+    await getPrismaClient().userHealthProfile.update({
+      where: { userId: user.id },
+      data: { aiIncludedSections: ["ABOUT_ME"] },
+    });
+    expect(await loadFreeWeekdaySpansForPrompt(user.id)).toEqual([]);
+    // The data surfaces still show the person their own answer.
+    expect(await loadFreeWeekdaySpans(user.id)).toHaveLength(1);
+  });
+});
+
 describe("AUDIT-C", () => {
   async function postAuditC(body: unknown): Promise<Response> {
     const { POST } = await import("@/app/api/anamnesis/audit-c/route");
@@ -396,6 +424,79 @@ describe("AUDIT-C", () => {
     });
   });
 
+  it("scores against the wording offered and refuses another one or a future date", async () => {
+    // An English account is shown the English wording; a German cut-off for
+    // it would be the stricter S3 table applied to questions never asked.
+    const man = await makeUser({ locale: "en", gender: "MALE" });
+    await signIn(man.id);
+    const wrong = await postAuditC({
+      items: [2, 1, 1],
+      locale: "de",
+      source: "IOS",
+    });
+    expect(wrong.status).toBe(422);
+    const { result } = await data<{
+      result: { locale: string; cutoff: { source: string } | null };
+    }>(await postAuditC({ items: [2, 1, 1], source: "IOS" }), 201);
+    expect(result.locale).toBe("en");
+    expect(result.cutoff?.source).toBe("BRADLEY_2007");
+
+    // French has no validated wording, so the English one is offered and
+    // stored, whatever the account language says.
+    const french = await makeUser({ locale: "fr", gender: "MALE" });
+    await signIn(french.id);
+    const fr = await data<{ result: { locale: string } }>(
+      await postAuditC({ items: [0, 0, 1], locale: "fr", source: "WEB" }),
+      201,
+    );
+    expect(fr.result.locale).toBe("en");
+
+    const future = await postAuditC({
+      items: [0, 0, 0],
+      source: "WEB",
+      takenAt: new Date(Date.now() + 60 * 60 * 1000).toISOString(),
+    });
+    expect(future.status).toBe(422);
+    const skew = await postAuditC({
+      items: [0, 0, 0],
+      source: "WEB",
+      takenAt: new Date(Date.now() + 60 * 1000).toISOString(),
+    });
+    expect(skew.status).toBe(201);
+  });
+
+  it("reads latest and the offer from the newest entry whose answers open", async () => {
+    const user = await makeUser();
+    await signIn(user.id);
+    const older = await data<{ result: { id: string } }>(
+      await postAuditC({
+        items: [1, 0, 0],
+        source: "WEB",
+        takenAt: "2025-01-10T09:00:00.000Z",
+      }),
+      201,
+    );
+    // A newer entry this host cannot open (its key is gone).
+    const sealed = await getPrismaClient().mentalHealthAssessment.create({
+      data: {
+        userId: user.id,
+        instrument: "AUDIT_C",
+        locale: "en",
+        responsesEncrypted: new Uint8Array([9, 9, 9, 9]),
+        totalScore: 11,
+        severityBand: "unbanded",
+        takenAt: new Date(),
+        source: "WEB",
+      },
+    });
+    const life = await readLife();
+    expect(life.auditC.latest?.id).toBe(older.result.id);
+    // Timed from the result shown, not from the entry nobody can read.
+    expect(life.auditC.offerDue).toBe(true);
+    // Still listed, so it can be removed.
+    expect(life.auditC.history.map((h) => h.id)).toContain(sealed.id);
+  });
+
   it("soft-deletes an entry once, and only an AUDIT-C entry of the record", async () => {
     const user = await makeUser();
     await signIn(user.id);
@@ -477,6 +578,53 @@ describe("AI switches and confirmations", () => {
     ]);
   });
 
+  it("never turns a section on again that a concurrent switch turned off", async () => {
+    const user = await makeUser({
+      modulePreferencesJson: { mentalHealth: true, coach: true },
+    });
+    await signIn(user.id);
+    const prisma = getPrismaClient();
+    await prisma.userHealthProfile.create({
+      data: { userId: user.id, aiIncludedSections: ["ABOUT_ME", "HABITS"] },
+    });
+    const { lockAiSections } = await import("@/lib/profile/life-circumstances");
+    const aboutMe = await import("@/app/api/coach/about-me/route");
+
+    // The switch-off holds its transaction open while the PUT starts.
+    let release!: () => void;
+    const held = new Promise<void>((resolve) => (release = resolve));
+    let locked!: () => void;
+    const isLocked = new Promise<void>((resolve) => (locked = resolve));
+    const switchOff = prisma.$transaction(async (tx) => {
+      await lockAiSections(tx, user.id);
+      await tx.userHealthProfile.update({
+        where: { userId: user.id },
+        data: { aiIncludedSections: ["ABOUT_ME"] },
+      });
+      locked();
+      await held;
+    });
+    await isLocked;
+    const put = (aboutMe.PUT as unknown as Handler)(
+      request("PUT", "/api/coach/about-me", {
+        aiIncludedSections: ["ABOUT_ME", "CONDITIONS"],
+      }),
+    );
+    // Long enough for an unlocked PUT to have read the old array.
+    await new Promise((resolve) => setTimeout(resolve, 300));
+    release();
+    await switchOff;
+    await data(await put);
+
+    const stored = await prisma.userHealthProfile.findUniqueOrThrow({
+      where: { userId: user.id },
+    });
+    expect([...stored.aiIncludedSections].sort()).toEqual([
+      "ABOUT_ME",
+      "CONDITIONS",
+    ]);
+  });
+
   it("answers a stale token with about_me_conflict", async () => {
     const user = await makeUser();
     await signIn(user.id);
diff --git a/tests/integration/pulse-day-mean-method.test.ts b/tests/integration/pulse-day-mean-method.test.ts
index 9aa1dff77..822b16dd1 100644
--- a/tests/integration/pulse-day-mean-method.test.ts
+++ b/tests/integration/pulse-day-mean-method.test.ts
@@ -38,7 +38,7 @@ import { readLiveBuckets } from "@/lib/measurements/daily-series-read";
 import { readCanonicalRollupBuckets } from "@/lib/rollups/measurement-read";
 import { recomputeUserRollups } from "@/lib/rollups/measurement-rollups";
 import { invalidateUserTimezone } from "@/lib/tz/resolver";
-import { pulseDayMeanDryRunSchema } from "@/lib/validations/pulse-day-mean";
+import { pulseDayMeanStoredRunsSchema } from "@/lib/validations/pulse-day-mean";
 import type { Job } from "pg-boss";
 
 const prisma = getPrismaClient();
@@ -276,7 +276,9 @@ describe("the dry run", () => {
     const stored = await prisma.appSettings.findUniqueOrThrow({
       where: { id: "singleton" },
     });
-    const run = pulseDayMeanDryRunSchema.parse(stored.pulseDayMeanDryRunJson);
+    const run = pulseDayMeanStoredRunsSchema.parse(
+      stored.pulseDayMeanDryRunJson,
+    );
     expect(run.status).toBe("done");
     expect(run.finishedAt).not.toBeNull();
     expect(run.result).toEqual({
@@ -324,6 +326,13 @@ describe("the rebuild after apply", () => {
     });
     const outcome = await handlePulseDayMean(job("rebuild"));
     expect(outcome.ok).toBe(true);
+    // The rebuild is on the record beside the dry run, with what it did.
+    const { readPulseDayMeanStatus } =
+      await import("@/lib/jobs/pulse-day-mean");
+    const status = await readPulseDayMeanStatus();
+    expect(status.lastRebuild).toMatchObject({ status: "done" });
+    expect(status.lastRebuild?.accounts).toBeGreaterThan(0);
+    expect(status.lastDryRun?.status).toBe("done");
 
     expect(await dayMean(A, D1)).toBeCloseTo(V2, 6);
     expect(await dayMean(A, D2)).toBeCloseTo(70, 6);
diff --git a/tests/integration/relation-links-portable-restore.test.ts b/tests/integration/relation-links-portable-restore.test.ts
new file mode 100644
index 000000000..308369684
--- /dev/null
+++ b/tests/integration/relation-links-portable-restore.test.ts
@@ -0,0 +1,229 @@
+/**
+ * The five relation link tables survive a PORTABLE export and restore.
+ *
+ * A portable file carries no row id for a medication or a lab result: the
+ * restore mints fresh ones, because the file may land on another instance or
+ * another account where the original ids mean nothing. The link tables name
+ * both of their ends by id, so a link that rode the file under the OLD id of
+ * either end used to point at nothing after the restore, and the restore's
+ * own placement check dropped it as "end is gone". Four of the five tables
+ * have a medication or a lab result at one end; only the visit to dose link
+ * came back.
+ *
+ * Driven through the real restore route. A portable file with documents is
+ * refused (no document bytes, see `backup-portable-file-restore.test.ts`), so
+ * the file restored here is the portable payload with the document entries
+ * taken from the disaster-recovery build of the same account: everything else
+ * (medications, lab results, links) is the portable bytes exactly as the
+ * export writes them.
+ *
+ * Mutation check: drop the `exportRef` from the portable medication or lab
+ * result entry and the matching link count goes to zero.
+ */
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+process.env.ENCRYPTION_KEY =
+  "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef";
+
+import { encrypt } from "@/lib/crypto";
+import { buildFullBackupPayload } from "@/lib/export/full-backup-payload";
+import { linkTargets } from "@/lib/links";
+import { parseBackupPayload } from "@/lib/validations/backup";
+import { POST } from "./restore-job-driver";
+
+import { cookieJar, headerJar } from "./mock-next-headers";
+import { getPrismaClient, truncateAllTables } from "./setup";
+
+vi.mock("next/headers", async () => {
+  const { cookieJar, headerJar } = await import("./mock-next-headers");
+  return {
+    headers: vi.fn(async () => ({
+      get: (name: string) => headerJar.get(name.toLowerCase()) ?? null,
+    })),
+    cookies: vi.fn(async () => ({
+      get: (name: string) => {
+        const value = cookieJar.get(name);
+        return value ? { name, value } : undefined;
+      },
+      set: (name: string, value: string) => cookieJar.set(name, value),
+      delete: (name: string) => cookieJar.delete(name),
+    })),
+  };
+});
+
+vi.mock("@/lib/db-compat", () => ({
+  ensureDbCompatibility: vi.fn().mockResolvedValue(undefined),
+}));
+
+vi.mock("@/lib/cache/invalidate", () => ({
+  invalidateUserData: vi.fn(),
+}));
+
+beforeEach(async () => {
+  await truncateAllTables(getPrismaClient());
+  cookieJar.clear();
+  headerJar.clear();
+});
+
+async function seed() {
+  const prisma = getPrismaClient();
+  const owner = await prisma.user.create({
+    data: {
+      username: "portable-links-owner",
+      email: "portable-links-owner@example.test",
+      role: "ADMIN",
+    },
+  });
+  const userId = owner.id;
+  const session = await prisma.session.create({
+    data: { userId, expiresAt: new Date(Date.now() + 60_000) },
+  });
+  cookieJar.set("healthlog_session", session.id);
+
+  const encounter = await prisma.encounter.create({
+    data: {
+      userId,
+      occurredAt: new Date("2026-07-01T09:00:00.000Z"),
+      status: "DONE",
+      kind: "ROUTINE",
+    },
+  });
+  const document = await prisma.inboundDocument.create({
+    data: {
+      userId,
+      kind: "LAB_RESULT",
+      title: "Thyroid panel",
+      mimeType: "application/pdf",
+      byteSize: 12,
+      contentEncrypted: new Uint8Array([1, 2, 3]),
+      contentCodec: "binary2",
+      documentDate: new Date("2026-06-30T00:00:00.000Z"),
+    },
+  });
+  const biomarker = await prisma.biomarker.create({
+    data: { userId, name: "TSH", unit: "mU/L" },
+  });
+  const labResult = await prisma.labResult.create({
+    data: {
+      userId,
+      biomarkerId: biomarker.id,
+      analyte: "TSH",
+      value: 2.1,
+      unit: "mU/L",
+      takenAt: new Date("2026-06-30T09:00:00.000Z"),
+    },
+  });
+  const episode = await prisma.illnessEpisode.create({
+    data: {
+      userId,
+      label: "Thyroid",
+      type: "CHRONIC",
+      onsetAt: new Date("2026-01-20T00:00:00.000Z"),
+    },
+  });
+  const medication = await prisma.medication.create({
+    data: { userId, name: "Levothyroxine", dose: "50 µg" },
+  });
+  // A second drug with no links, so a restore that bound every link to the
+  // first medication it found would be told apart from one that maps by key.
+  await prisma.medication.create({
+    data: { userId, name: "Vitamin D", dose: "1000 IU" },
+  });
+  const vaccination = await prisma.vaccinationRecord.create({
+    data: {
+      userId,
+      occurredAt: new Date("2026-05-14T00:00:00.000Z"),
+      antigenSlug: "tdap",
+      vaccineName: "Tdap",
+    },
+  });
+
+  const link = (
+    sourceKind: Parameters<typeof linkTargets>[1]["sourceKind"],
+    sourceId: string,
+    targetKind: Parameters<typeof linkTargets>[1]["targetKind"],
+    targetId: string,
+  ) =>
+    linkTargets(prisma, {
+      userId,
+      sourceKind,
+      sourceId,
+      targetKind,
+      targetIds: [targetId],
+    });
+  await link("labResult", labResult.id, "document", document.id);
+  await link("encounter", encounter.id, "medication", medication.id);
+  await link("encounter", encounter.id, "vaccination", vaccination.id);
+  await link("medication", medication.id, "document", document.id);
+  await link("medication", medication.id, "conditionEpisode", episode.id);
+
+  return { userId, medication, labResult };
+}
+
+describe("a portable restore of the five relation link tables", () => {
+  it("brings every link back, bound to the records the restore wrote", async () => {
+    const prisma = getPrismaClient();
+    const { userId, medication, labResult } = await seed();
+
+    const portable = parseBackupPayload(
+      JSON.stringify((await buildFullBackupPayload(prisma, userId)).payload),
+    );
+    const dr = parseBackupPayload(
+      JSON.stringify(
+        (
+          await buildFullBackupPayload(prisma, userId, {
+            purpose: "disaster-recovery",
+          })
+        ).payload,
+      ),
+    );
+    // The premise: the portable file names no medication or lab result id.
+    expect(portable.medications.every((m) => m.id === undefined)).toBe(true);
+    expect(portable.labResults.every((l) => l.id === undefined)).toBe(true);
+
+    const file = parseBackupPayload(
+      JSON.stringify({ ...portable, documents: dr.documents }),
+    );
+    const backup = await prisma.dataBackup.create({
+      data: {
+        userId,
+        type: "PORTABLE_LINKS_RESTORE",
+        data: encrypt(JSON.stringify(file)),
+      },
+    });
+
+    const response = await POST(
+      new Request(`http://localhost/api/admin/backups/${backup.id}/restore`, {
+        method: "POST",
+        headers: { "content-type": "application/json" },
+        body: JSON.stringify({ confirm: "RESTORE" }),
+      }) as unknown as Parameters<typeof POST>[0],
+      { params: Promise.resolve({ id: backup.id }) },
+    );
+    expect(response.status, await response.clone().text()).toBe(200);
+
+    // The ends were re-minted, so the old ids are gone.
+    const restoredMed = await prisma.medication.findFirstOrThrow({
+      where: { userId, name: "Levothyroxine" },
+    });
+    const restoredLab = await prisma.labResult.findFirstOrThrow({
+      where: { userId, analyte: "TSH" },
+    });
+    expect(restoredMed.id).not.toBe(medication.id);
+    expect(restoredLab.id).not.toBe(labResult.id);
+
+    const where = { userId };
+    const [labDocs, encMeds, encDoses, medDocs, medConds] = await Promise.all([
+      prisma.labResultDocumentLink.findMany({ where }),
+      prisma.encounterMedicationLink.findMany({ where }),
+      prisma.encounterVaccinationLink.findMany({ where }),
+      prisma.medicationDocumentLink.findMany({ where }),
+      prisma.medicationConditionLink.findMany({ where }),
+    ]);
+    expect(labDocs.map((r) => r.labResultId)).toEqual([restoredLab.id]);
+    expect(encMeds.map((r) => r.medicationId)).toEqual([restoredMed.id]);
+    expect(encDoses).toHaveLength(1);
+    expect(medDocs.map((r) => r.medicationId)).toEqual([restoredMed.id]);
+    expect(medConds.map((r) => r.medicationId)).toEqual([restoredMed.id]);
+  });
+});