fix(cron): avoid caching deadline-clamped probes

Cleo Thornsburg · Cleo Thornsburg · commit 9a8fd3de8785 · 2026-06-15T10:49:18.000-05:00
diff --git a/src/cron/isolated-agent/model-preflight.runtime.test.ts b/src/cron/isolated-agent/model-preflight.runtime.test.ts
@@ -270,6 +270,44 @@ describe("preflightCronModelProvider", () => {
     expect(fetchWithSsrFGuardMock).toHaveBeenCalledTimes(2);
   });
 
+  it("does not cache a failure from a deadline-clamped probe", async () => {
+    vi.useFakeTimers();
+    vi.setSystemTime(1_000);
+    fetchWithSsrFGuardMock.mockRejectedValueOnce(new DOMException("timed out", "TimeoutError"));
+
+    const cfg = {
+      models: {
+        providers: {
+          ollama: {
+            api: "ollama" as const,
+            baseUrl: "http://127.0.0.1:11434",
+            models: [],
+          },
+        },
+      },
+    };
+
+    const first = await preflightCronModelProvider({
+      cfg,
+      provider: "ollama",
+      model: "qwen3:14b",
+      deadlineMs: 1_200,
+    });
+
+    expect(first.status).toBe("unavailable");
+    expect(requireFetchPreflightRequest().timeoutMs).toBe(200);
+
+    mockReachableResponse(200);
+    const nextRun = await preflightCronModelProvider({
+      cfg,
+      provider: "ollama",
+      model: "qwen3:14b",
+    });
+
+    expect(nextRun).toEqual({ status: "available" });
+    expect(fetchWithSsrFGuardMock).toHaveBeenCalledTimes(2);
+  });
+
   it("retries an unavailable endpoint after the cache ttl", async () => {
     fetchWithSsrFGuardMock.mockRejectedValueOnce(new Error("ECONNREFUSED")).mockResolvedValueOnce({
       response: { status: 200 },
diff --git a/src/cron/isolated-agent/model-preflight.runtime.ts b/src/cron/isolated-agent/model-preflight.runtime.ts
@@ -263,6 +263,7 @@ export async function preflightCronModelProvider(params: {
   let lastError: unknown;
   let attempts = 0;
   let budgetExhausted = false;
+  let cacheableFailure = true;
   for (let attempt = 1; attempt <= maxAttempts; attempt += 1) {
     const remainingBudgetMs = resolveRemainingBudgetMs(params.deadlineMs);
     if (remainingBudgetMs !== undefined && remainingBudgetMs <= 0) {
@@ -271,18 +272,24 @@ export async function preflightCronModelProvider(params: {
       break;
     }
     attempts = attempt;
+    const probeTimeoutMs =
+      remainingBudgetMs === undefined ? timeoutMs : Math.min(timeoutMs, remainingBudgetMs);
     try {
       await probeLocalProviderEndpoint({
         api,
         baseUrl,
-        timeoutMs:
-          remainingBudgetMs === undefined ? timeoutMs : Math.min(timeoutMs, remainingBudgetMs),
+        timeoutMs: probeTimeoutMs,
       });
       const result: EndpointPreflightResult = { status: "available" };
       preflightCache.set(cacheKey, { checkedAtMs: nowMs, result });
       return { status: "available" };
     } catch (error) {
       lastError = error;
+      // A deadline-clamped probe did not receive the configured health-check
+      // window, so its failure must not poison the endpoint cache.
+      if (probeTimeoutMs < timeoutMs) {
+        cacheableFailure = false;
+      }
       if (attempt < maxAttempts) {
         const remainingDelayBudgetMs = resolveRemainingBudgetMs(params.deadlineMs);
         if (remainingDelayBudgetMs !== undefined && remainingDelayBudgetMs <= 0) {
@@ -305,7 +312,7 @@ export async function preflightCronModelProvider(params: {
     error: lastError ?? new Error("cron model preflight chain budget exhausted"),
     attempts,
   };
-  if (!budgetExhausted) {
+  if (!budgetExhausted && cacheableFailure) {
     preflightCache.set(cacheKey, { checkedAtMs: nowMs, result });
   }
   return buildUnavailableResult({
