test(core,webapp): terminate in-process agent e2e runs reliably

The in-process chat.agent e2e harness runs every leg against the single
globalThis API registry. When a leg closed while its run was still
parked on a session waitpoint, that run stayed alive into the next leg
and its later teardown unregistered the globals out from under the
active run. In CI this surfaced as Noop run-metadata and "can only be
used from inside task.run()" errors across most legs.

close() now aborts the run and awaits full teardown instead of racing a
10s timeout; the abort disables the session-waitpoint backend so any
in-flight or later wait() returns at once; and the backend keeps its
pending entry until the wait resolves so disable() can abort it. Runs
no longer overlap, so the shared globals stay consistent.

Author: ericallam

apps/webapp/test/helpers/agentHarness.ts

@@ -61,7 +61,7 @@ export function runRealChatAgent(opts: RunRealChatAgentOptions): RunningAgent {
   });
   const apiClient = apiClientManager.clientOrThrow();
   const manager = new StandardSessionStreamManager(apiClient, opts.baseUrl);
-  const { runtimeManager, restore } = installSessionWaitpointBackend(apiClient);
+  const { backend, runtimeManager, restore } = installSessionWaitpointBackend(apiClient);
 
   const taskEntry = resourceCatalog.getTask(opts.agentId);
   if (!taskEntry) {
@@ -74,6 +74,11 @@ export function runRealChatAgent(opts: RunRealChatAgentOptions): RunningAgent {
   ) => Promise<unknown>;
 
   const runSignal = new AbortController();
+  runSignal.signal.addEventListener("abort", () => {
+    try {
+      backend.disable();
+    } catch {}
+  });
   const runId = opts.runId ?? `run_${opts.addressingKey}`;
 
   const idle =
@@ -128,10 +133,7 @@ export function runRealChatAgent(opts: RunRealChatAgentOptions): RunningAgent {
         );
       } catch {}
       runSignal.abort();
-      await Promise.race([
-        done.catch(() => {}),
-        new Promise((resolve) => setTimeout(resolve, 10_000)),
-      ]);
+      await done.catch(() => {});
     },
   };
 }

packages/core/src/v3/test/session-waitpoint-backend.ts

@@ -58,6 +58,7 @@ function parseTimeoutMs(timeout: string | undefined): number | undefined {
  */
 export class SessionWaitpointBackend {
   private readonly pending = new Map<string, PendingWait>();
+  private disabled = false;
 
   constructor(private readonly apiClient: ApiClient) {}
 
@@ -76,11 +77,17 @@ export class SessionWaitpointBackend {
   }
 
   async wait(waitpointFriendlyId: string): Promise<WaitpointTokenResult> {
+    if (this.disabled) {
+      return {
+        ok: false,
+        output: JSON.stringify({ message: "Session waitpoint backend disabled" }),
+        outputType: "application/json",
+      };
+    }
     const pending = this.pending.get(waitpointFriendlyId);
     if (!pending) {
       return { ok: true };
     }
-    this.pending.delete(waitpointFriendlyId);
 
     const timeoutMs = parseTimeoutMs(pending.timeout);
     let timer: ReturnType<typeof setTimeout> | undefined;
@@ -118,10 +125,12 @@ export class SessionWaitpointBackend {
       if (timer) {
         clearTimeout(timer);
       }
+      this.pending.delete(waitpointFriendlyId);
     }
   }
 
   disable(): void {
+    this.disabled = true;
     for (const pending of this.pending.values()) {
       pending.abort.abort();
     }