fix(gateway): log start session persistence failures

masatohoshino · masatohoshino · commit e8f4c93c6643 · 2026-07-05T21:42:50.000+09:00
The gateway session-lifecycle "start" event persistence (persistGatewaySessionLifecycleEvent, which surfaces write failures via requireWriteSuccess) was fired as void ...catch(() => undefined), swallowing the rejection with zero logging. A failed start-marker write silently dropped the run's start record from restart-recovery accounting, with no operator-visible trace. The sibling terminal-phase catch already logs this since #97839; the start path was the unfixed sibling. Mirror that fix: log the swallowed start-phase persistence failure with the same redacted message shape via formatForLog, keeping the fire-and-forget semantics unchanged. Adds a focused regression test asserting the log fires on start-persist rejection.
diff --git a/src/gateway/server-chat.agent-events.test.ts b/src/gateway/server-chat.agent-events.test.ts
@@ -3174,6 +3174,32 @@ describe("agent event handler", () => {
     );
   });
 
+  it("logs when start session persistence fails", async () => {
+    const { chatRunState, handler, sessionEventSubscribers } = createHarness();
+    sessionEventSubscribers.subscribe("conn-1");
+    chatRunState.registry.add("run-global-work", {
+      sessionKey: "global",
+      agentId: "work",
+      clientRunId: "client-global-work",
+    });
+    persistGatewaySessionLifecycleEventMock.mockRejectedValueOnce(new Error("start disk full"));
+
+    handler({
+      runId: "run-global-work",
+      seq: 1,
+      stream: "lifecycle",
+      ts: Date.now(),
+      data: { phase: "start" },
+    });
+
+    await vi.waitFor(() => {
+      expect(logErrorMock).toHaveBeenCalledTimes(1);
+    });
+    expect(logErrorMock).toHaveBeenCalledWith(
+      "gateway: start session persistence failed session=global run=run-global-work error=Error: start disk full",
+    );
+  });
+
   it("routes hidden selected-agent global chat events only to matching subscribers", () => {
     const { broadcastToConnIds, chatRunState, handler, sessionMessageSubscribers } =
       createHarness();
diff --git a/src/gateway/server-chat.ts b/src/gateway/server-chat.ts
@@ -1439,7 +1439,14 @@ export function createAgentEventHandler({
         sessionKey,
         agentId: sessionAgentId,
         event: evt,
-      }).catch(() => undefined);
+      }).catch((err: unknown) => {
+        // Surface the swallowed start-phase persistence failure: a silent write
+        // failure drops the run's start marker from restart-recovery accounting
+        // with no operator trace, matching the terminal-phase log below.
+        logError(
+          `gateway: start session persistence failed session=${formatForLog(sessionKey)} run=${formatForLog(evt.runId)} error=${formatForLog(err)}`,
+        );
+      });
       const sessionEventConnIds = sessionEventSubscribers.getAll();
       if (sessionEventConnIds.size > 0) {
         broadcastToConnIds(
