fix(release): stabilize config restart QA

steipete · steipete · commit cc91ff04ccac · 2026-05-22T15:53:50.000+01:00
diff --git a/qa/scenarios/config/config-restart-capability-flip.md b/qa/scenarios/config/config-restart-capability-flip.md
@@ -66,6 +66,9 @@ steps:
           - ref: env
           - Capability flip
           - ref: sessionKey
+      - set: setupStartIndex
+        value:
+          expr: state.getSnapshot().messages.length
       - try:
           actions:
             - call: patchConfig
@@ -92,6 +95,15 @@ steps:
                     expr: config.setupPrompt
                   timeoutMs:
                     expr: liveTurnTimeoutMs(env, 30000)
+            - call: waitForOutboundMessage
+              args:
+                - ref: state
+                - lambda:
+                    params: [candidate]
+                    expr: "candidate.conversation.id === 'qa-operator' && String(candidate.text ?? '').includes('Protocol note')"
+                - expr: liveTurnTimeoutMs(env, 30000)
+                - sinceIndex:
+                    ref: setupStartIndex
             - call: readEffectiveTools
               saveAs: beforeTools
               args:
diff --git a/src/agents/session-write-lock.test.ts b/src/agents/session-write-lock.test.ts
@@ -1,3 +1,4 @@
+import fsSync from "node:fs";
 import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
@@ -763,6 +764,35 @@ describe("acquireSessionWriteLock", () => {
     }
   });
 
+  it("retries when a reported stale same-process lock disappears before recovery", async () => {
+    await withTempSessionLockFile(async ({ sessionFile, lockPath }) => {
+      await fs.writeFile(
+        lockPath,
+        JSON.stringify({
+          pid: process.pid,
+          createdAt: new Date().toISOString(),
+          starttime: FAKE_STARTTIME,
+        }),
+        "utf8",
+      );
+      let resolverCalls = 0;
+      testing.setProcessStartTimeResolverForTest((pid) => {
+        if (pid !== process.pid) {
+          return null;
+        }
+        resolverCalls += 1;
+        if (resolverCalls === 1) {
+          fsSync.rmSync(lockPath, { force: true });
+        }
+        return FAKE_STARTTIME;
+      });
+
+      const lock = await acquireSessionWriteLock({ sessionFile, timeoutMs: 500 });
+      await lock.release();
+      expect(resolverCalls).toBeGreaterThan(0);
+    });
+  });
+
   it("removes held locks on termination signals", async () => {
     const signals = ["SIGINT", "SIGTERM", "SIGQUIT", "SIGABRT"] as const;
     const originalKill = process.kill.bind(process);
diff --git a/src/agents/session-write-lock.ts b/src/agents/session-write-lock.ts
@@ -560,6 +560,16 @@ async function removeReportedStaleLockIfStillStale(params: {
 }): Promise<boolean> {
   const nowMs = Date.now();
   const payload = await readLockPayload(params.lockPath);
+  if (payload === null) {
+    try {
+      await fs.access(params.lockPath);
+    } catch (error) {
+      if ((error as NodeJS.ErrnoException).code === "ENOENT") {
+        return true;
+      }
+      throw error;
+    }
+  }
   const inspected = inspectLockPayloadForSession({
     payload,
     staleMs: params.staleMs,
