diff --git a/.planning/phase-W-INSIGHTS-HOT-report.md b/.planning/phase-W-INSIGHTS-HOT-report.md
new file mode 100644
index 000000000..9f3984893
--- /dev/null
+++ b/.planning/phase-W-INSIGHTS-HOT-report.md
@@ -0,0 +1,105 @@
+# Phase report — W-INSIGHTS-HOT (v1.4.41)
+
+## Mission
+
+Eliminate the recurring ~14 s warm response on the two iOS-facing insight
+endpoints:
+
+- `GET /api/insights/blood-pressure-status`
+- `GET /api/insights/weight-status`
+
+iOS v0.5.4 hits both on every dashboard open. Marc feels them daily.
+
+## Root cause
+
+The sibling `bmi-status` route shipped a **timeout-stub persist** in
+v1.4.37: when the AI provider stalls past the 20 s budget, the route
+writes a `model: "timeout-stub", timeout: true` sentinel row keyed to
+today's Berlin day. The next mount hits the cache-lookup branch and
+short-circuits — no further provider race that day.
+
+`blood-pressure-status` and `weight-status` carried only the bare-fallback
+return (no persist), so every reload re-raced the same 20 s provider
+call. That is the ~14 s warm response Marc sees (the provider call
+itself averages ~14 s on cold paths before the 20 s timeout cuts it).
+
+## Fix
+
+Mirror the bmi-status stub-persist pattern verbatim into both routes.
+
+- `src/lib/insights/blood-pressure-status.ts` — lines 544-551
+  (pre-existing return) replaced with the 40-line stub-persist block
+  (best-effort `prisma.auditLog.create` then return the stub envelope).
+- `src/lib/insights/weight-status.ts` — same shape, same block.
+
+Response shape unchanged. iOS v0.5.4 contract preserved:
+`{ hasProvider, text, cached, updatedAt }`.
+
+## Files touched
+
+- `src/lib/insights/blood-pressure-status.ts` (+43 / −4)
+- `src/lib/insights/weight-status.ts` (+43 / −4)
+- `src/lib/insights/__tests__/blood-pressure-status.test.ts` (+98)
+- `src/lib/insights/__tests__/weight-status.test.ts` (+86)
+
+## Tests added
+
+Two new test suites per route, copied from the bmi-status template:
+
+1. **Timeout-stub persistence** — provider hangs past timeout, asserts
+   one `auditLog.create` call with `timeout: true`, `model:
+   "timeout-stub"`, `dateKey` matching `YYYY-MM-DD`.
+2. **Cache short-circuit** — second mount with a stub row in
+   `findFirst` returns the cached text, never calls the provider, never
+   writes a new row.
+
+Both use `vi.useFakeTimers()` + `advanceTimersByTimeAsync(25_000)` so
+the suite does not actually wait 20 s.
+
+## Verification
+
+- `pnpm test src/lib/insights` — 148/148 pass (up from 144).
+- `pnpm typecheck` — clean (after `pnpm db:generate` to materialise the
+  Prisma client in the worktree).
+- `pnpm lint` — clean (only pre-existing warnings in `insights/page.tsx`
+  and `summaries-slice.ts`, unrelated).
+
+## Perf delta (estimated)
+
+| State | Before | After |
+|---|---|---|
+| Cache hit (fresh insight or stub) | ~50 ms | ~50 ms (unchanged) |
+| Provider available, no cache | ~3-5 s (real AI call) | ~3-5 s (unchanged) |
+| Provider stall, no cache, **first** mount of the day | ~20 s timeout | ~20 s timeout (unchanged) |
+| Provider stall, no cache, **subsequent** mounts | **~14-20 s every time** | **~50 ms (cache hit on stub)** |
+
+Marc's reported ~14 s recurring warm response is the **subsequent-mount**
+case — that is the row the fix eliminates. The first daily stall still
+takes the 20 s budget, same as bmi-status.
+
+## Risk callouts for reconcile
+
+1. **Stub-persist comment carries forward bmi-status' optimistic note**
+   about the pre-warm worker overwriting the stub. In reality the
+   pre-warm worker uses `force: false`, so it also short-circuits on
+   the stub. This is consistent with bmi-status' shipped behaviour and
+   matches the day-key invalidation cycle (the stub naturally clears at
+   midnight Berlin time when the dateKey rolls). No code change needed;
+   noting it so the reconcile reviewer does not re-open the question.
+2. **No shape change** — the response envelope, the `auditLog` row
+   shape, and the cache-lookup branch are unchanged. No iOS or
+   web-frontend touch required.
+3. **No new dependencies, no provider-side changes, no DB migration.**
+4. **The two test suites use fake timers**, mirroring the bmi-status
+   pattern that has been stable since v1.4.37 — should be safe under
+   the existing vitest config.
+
+## Commits
+
+Single atomic commit on `worktree-agent-a9d3a2cf26b93d03d`:
+
+```
+perf(insights): persist timeout stub for blood-pressure and weight status
+```
+
+Pushed to origin at end of session.
diff --git a/src/lib/insights/__tests__/blood-pressure-status.test.ts b/src/lib/insights/__tests__/blood-pressure-status.test.ts
index 7b089aaec..aa7f9bfa9 100644
--- a/src/lib/insights/__tests__/blood-pressure-status.test.ts
+++ b/src/lib/insights/__tests__/blood-pressure-status.test.ts
@@ -109,6 +109,104 @@ describe("generateBloodPressureStatusForUser — v1.4.6 bucketed payload", () =>
   });
 });
 
+describe("generateBloodPressureStatusForUser — v1.4.41 timeout-stub persistence", () => {
+  it("persists a sentinel row keyed to today when the provider times out", async () => {
+    vi.mocked(prisma.user.findUnique).mockResolvedValue({
+      dateOfBirth: null,
+    } as never);
+    vi.mocked(prisma.auditLog.findFirst).mockResolvedValue(null);
+    vi.mocked(prisma.measurement.findMany).mockResolvedValue([
+      { type: "BLOOD_PRESSURE_SYS", value: 132, measuredAt: new Date() },
+      { type: "BLOOD_PRESSURE_DIA", value: 84, measuredAt: new Date() },
+    ] as never);
+    vi.mocked(prisma.medication.findMany).mockResolvedValue([] as never);
+    vi.mocked(prisma.medicationIntakeEvent.findMany).mockResolvedValue(
+      [] as never,
+    );
+    vi.mocked(prisma.moodEntry.findMany).mockResolvedValue([] as never);
+    vi.mocked(prisma.auditLog.create).mockResolvedValue({
+      createdAt: new Date(),
+    } as never);
+
+    // Provider hangs past the 20 s timeout — withTimeout cuts the race.
+    vi.mocked(resolveProvider).mockResolvedValue({
+      type: "anthropic",
+      generateCompletion: vi.fn(
+        () =>
+          new Promise(() => {
+            /* never resolves */
+          }),
+      ),
+    } as never);
+
+    vi.useFakeTimers();
+    const promise = generateBloodPressureStatusForUser("user-1", {
+      locale: "en",
+    });
+    await vi.advanceTimersByTimeAsync(25_000);
+    const result = await promise;
+    vi.useRealTimers();
+
+    expect(result.text).toBeTruthy();
+    expect(result.cached).toBe(true);
+    expect(result.updatedAt).toBeTruthy();
+
+    const createCalls = vi.mocked(prisma.auditLog.create).mock.calls;
+    expect(createCalls.length).toBe(1);
+    const details = (createCalls[0][0] as { data: { details: string } }).data
+      .details;
+    const parsed = JSON.parse(details) as {
+      dateKey?: string;
+      text?: string;
+      timeout?: boolean;
+      model?: string;
+    };
+    expect(parsed.text).toBeTruthy();
+    expect(parsed.timeout).toBe(true);
+    expect(parsed.model).toBe("timeout-stub");
+    expect(parsed.dateKey).toMatch(/^\d{4}-\d{2}-\d{2}$/);
+  });
+
+  it("subsequent mounts short-circuit at the cache lookup and skip the race", async () => {
+    const parts = new Intl.DateTimeFormat("en-US", {
+      timeZone: "Europe/Berlin",
+      year: "numeric",
+      month: "2-digit",
+      day: "2-digit",
+    }).formatToParts(new Date());
+    const y = parts.find((p) => p.type === "year")!.value;
+    const m = parts.find((p) => p.type === "month")!.value;
+    const d = parts.find((p) => p.type === "day")!.value;
+    const todayKey = `${y}-${m}-${d}`;
+    const stubRow = {
+      createdAt: new Date(),
+      details: JSON.stringify({
+        dateKey: todayKey,
+        locale: "en",
+        text: "Blood-pressure assessment temporarily unavailable.",
+        timeout: true,
+      }),
+    };
+
+    vi.mocked(prisma.auditLog.findFirst).mockResolvedValue(stubRow as never);
+    const providerCall = vi.fn();
+    vi.mocked(resolveProvider).mockResolvedValue({
+      type: "anthropic",
+      generateCompletion: providerCall,
+    } as never);
+
+    const result = await generateBloodPressureStatusForUser("user-1", {
+      locale: "en",
+    });
+
+    expect(providerCall).not.toHaveBeenCalled();
+    expect(result.text).toBe("Blood-pressure assessment temporarily unavailable.");
+    expect(result.cached).toBe(true);
+    expect(result.updatedAt).toBe(stubRow.createdAt.toISOString());
+    expect(prisma.auditLog.create).not.toHaveBeenCalled();
+  });
+});
+
 describe("generateBloodPressureStatusForUser — token-leak hardening (v1.4.27 F16)", () => {
   it("strips metric: tokens out of the cached text before persisting", async () => {
     vi.mocked(prisma.user.findUnique).mockResolvedValue({
diff --git a/src/lib/insights/__tests__/weight-status.test.ts b/src/lib/insights/__tests__/weight-status.test.ts
index 82409970f..858aee7f7 100644
--- a/src/lib/insights/__tests__/weight-status.test.ts
+++ b/src/lib/insights/__tests__/weight-status.test.ts
@@ -82,6 +82,92 @@ describe("generateWeightStatusForUser — v1.4.6 bucketed payload", () => {
   });
 });
 
+describe("generateWeightStatusForUser — v1.4.41 timeout-stub persistence", () => {
+  it("persists a sentinel row keyed to today when the provider times out", async () => {
+    vi.mocked(prisma.auditLog.findFirst).mockResolvedValue(null);
+    vi.mocked(prisma.measurement.findMany).mockResolvedValue([
+      { type: "WEIGHT", value: 82, measuredAt: new Date() },
+    ] as never);
+    vi.mocked(prisma.moodEntry.findMany).mockResolvedValue([] as never);
+    vi.mocked(prisma.auditLog.create).mockResolvedValue({
+      createdAt: new Date(),
+    } as never);
+
+    // Provider hangs past the 20 s timeout — withTimeout cuts the race.
+    vi.mocked(resolveProvider).mockResolvedValue({
+      type: "anthropic",
+      generateCompletion: vi.fn(
+        () =>
+          new Promise(() => {
+            /* never resolves */
+          }),
+      ),
+    } as never);
+
+    vi.useFakeTimers();
+    const promise = generateWeightStatusForUser("user-1", { locale: "en" });
+    await vi.advanceTimersByTimeAsync(25_000);
+    const result = await promise;
+    vi.useRealTimers();
+
+    expect(result.text).toBeTruthy();
+    expect(result.cached).toBe(true);
+    expect(result.updatedAt).toBeTruthy();
+
+    const createCalls = vi.mocked(prisma.auditLog.create).mock.calls;
+    expect(createCalls.length).toBe(1);
+    const details = (createCalls[0][0] as { data: { details: string } }).data
+      .details;
+    const parsed = JSON.parse(details) as {
+      dateKey?: string;
+      text?: string;
+      timeout?: boolean;
+      model?: string;
+    };
+    expect(parsed.text).toBeTruthy();
+    expect(parsed.timeout).toBe(true);
+    expect(parsed.model).toBe("timeout-stub");
+    expect(parsed.dateKey).toMatch(/^\d{4}-\d{2}-\d{2}$/);
+  });
+
+  it("subsequent mounts short-circuit at the cache lookup and skip the race", async () => {
+    const parts = new Intl.DateTimeFormat("en-US", {
+      timeZone: "Europe/Berlin",
+      year: "numeric",
+      month: "2-digit",
+      day: "2-digit",
+    }).formatToParts(new Date());
+    const y = parts.find((p) => p.type === "year")!.value;
+    const m = parts.find((p) => p.type === "month")!.value;
+    const d = parts.find((p) => p.type === "day")!.value;
+    const todayKey = `${y}-${m}-${d}`;
+    const stubRow = {
+      createdAt: new Date(),
+      details: JSON.stringify({
+        dateKey: todayKey,
+        locale: "en",
+        text: "Weight assessment temporarily unavailable.",
+        timeout: true,
+      }),
+    };
+
+    vi.mocked(prisma.auditLog.findFirst).mockResolvedValue(stubRow as never);
+    const providerCall = vi.fn();
+    vi.mocked(resolveProvider).mockResolvedValue({
+      type: "anthropic",
+      generateCompletion: providerCall,
+    } as never);
+
+    const result = await generateWeightStatusForUser("user-1", { locale: "en" });
+
+    expect(providerCall).not.toHaveBeenCalled();
+    expect(result.text).toBe("Weight assessment temporarily unavailable.");
+    expect(result.cached).toBe(true);
+    expect(result.updatedAt).toBe(stubRow.createdAt.toISOString());
+    expect(prisma.auditLog.create).not.toHaveBeenCalled();
+  });
+});
+
 describe("generateWeightStatusForUser — token-leak hardening (v1.4.27 F16)", () => {
   it("strips metric: tokens out of the cached text before persisting", async () => {
     vi.mocked(prisma.auditLog.findFirst).mockResolvedValue(null);
diff --git a/src/lib/insights/blood-pressure-status.ts b/src/lib/insights/blood-pressure-status.ts
index c7f03e84c..009927bfa 100644
--- a/src/lib/insights/blood-pressure-status.ts
+++ b/src/lib/insights/blood-pressure-status.ts
@@ -542,11 +542,50 @@ export async function generateBloodPressureStatusForUser(
   );
 
   if (raced.timedOut || raced.value === null) {
+    // v1.4.41 — parity with bmi-status: persist a sentinel row keyed
+    // to today so the next mount short-circuits at the cache lookup
+    // above instead of re-racing the same 20 s provider call on every
+    // cold visit. Without this, a single stall (provider hiccup,
+    // network blip, model warm-up) left iOS users staring at the
+    // loading state on every dashboard open for the rest of the day.
+    // The body is the same no-key fallback the caller returned
+    // before, so the user-facing UI is identical; only the re-fire
+    // frequency drops to zero until the daily pre-warm job overwrites
+    // the stub. `meta.timeout` is set so the pre-warm worker can
+    // recognise and replace it rather than respect it as a real
+    // assessment.
+    const stubText = getNoKeyBloodPressureStatusText(locale);
+    let stubUpdatedAt: string | null = null;
+    try {
+      const stub = await prisma.auditLog.create({
+        data: {
+          userId,
+          action: cacheAction,
+          details: JSON.stringify({
+            dateKey: todayKey,
+            locale,
+            text: stubText,
+            providerType: provider.type,
+            model: "timeout-stub",
+            tokensUsed: null,
+            timeout: true,
+          }),
+        },
+        select: { createdAt: true },
+      });
+      stubUpdatedAt = stub.createdAt.toISOString();
+    } catch {
+      // Best-effort persist — if the row write fails the caller still
+      // sees the deterministic fallback text and the next mount falls
+      // back to the (still expensive) race. Do not surface the error:
+      // the user does not care that the cache write missed, only that
+      // the page rendered.
+    }
     return {
       hasProvider: true,
-      text: getNoKeyBloodPressureStatusText(locale),
+      text: stubText,
       cached: true,
-      updatedAt: null,
+      updatedAt: stubUpdatedAt,
     };
   }
 
diff --git a/src/lib/insights/weight-status.ts b/src/lib/insights/weight-status.ts
index 4f0a72adf..77f7a0351 100644
--- a/src/lib/insights/weight-status.ts
+++ b/src/lib/insights/weight-status.ts
@@ -440,11 +440,50 @@ export async function generateWeightStatusForUser(
   );
 
   if (raced.timedOut || raced.value === null) {
+    // v1.4.41 — parity with bmi-status: persist a sentinel row keyed
+    // to today so the next mount short-circuits at the cache lookup
+    // above instead of re-racing the same 20 s provider call on every
+    // cold visit. Without this, a single stall (provider hiccup,
+    // network blip, model warm-up) left iOS users staring at the
+    // loading state on every dashboard open for the rest of the day.
+    // The body is the same no-key fallback the caller returned
+    // before, so the user-facing UI is identical; only the re-fire
+    // frequency drops to zero until the daily pre-warm job overwrites
+    // the stub. `meta.timeout` is set so the pre-warm worker can
+    // recognise and replace it rather than respect it as a real
+    // assessment.
+    const stubText = getNoKeyWeightStatusText(locale);
+    let stubUpdatedAt: string | null = null;
+    try {
+      const stub = await prisma.auditLog.create({
+        data: {
+          userId,
+          action: cacheAction,
+          details: JSON.stringify({
+            dateKey: todayKey,
+            locale,
+            text: stubText,
+            providerType: provider.type,
+            model: "timeout-stub",
+            tokensUsed: null,
+            timeout: true,
+          }),
+        },
+        select: { createdAt: true },
+      });
+      stubUpdatedAt = stub.createdAt.toISOString();
+    } catch {
+      // Best-effort persist — if the row write fails the caller still
+      // sees the deterministic fallback text and the next mount falls
+      // back to the (still expensive) race. Do not surface the error:
+      // the user does not care that the cache write missed, only that
+      // the page rendered.
+    }
     return {
       hasProvider: true,
-      text: getNoKeyWeightStatusText(locale),
+      text: stubText,
       cached: true,
-      updatedAt: null,
+      updatedAt: stubUpdatedAt,
     };
   }