Stabilize Node CopilotRequestHandler e2e: align idle timeout with vitest budget

session.sendAndWait defaults its session.idle wait to 60s, but these e2e tests
declare a 90s vitest budget. The lower internal default became the binding
constraint, so under heavy parallel CI load the full runtime spawn + WebSocket
handshake + inference occasionally crossed 60s (but stayed under 90s) and failed
with "Timeout after 60000ms waiting for session.idle" instead of using the
budget the tests intended.

Pass an explicit 80s idle timeout to each sendAndWait (headroom left for the
finally-block disconnect within the 90s budget) so the effective budget matches
the declared one.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: SteveSandersonMS

nodejs/test/e2e/copilot_request_cancel_error.e2e.test.ts

@@ -143,8 +143,11 @@ describe("CopilotRequestHandler surfaces inference errors", async () => {
         const session = await client.createSession({ onPermissionRequest: approveAll });
         try {
             // The callback throws on inference; the turn surfaces an error (or
-            // completes without an assistant message) rather than hanging.
-            await session.sendAndWait({ prompt: "Say OK." }).catch(() => undefined);
+            // completes without an assistant message) rather than hanging. Use
+            // an explicit idle timeout aligned with the 90s vitest budget so a
+            // slow runtime spawn under load doesn't fire the lower (60s) default
+            // before the inference callback is even reached.
+            await session.sendAndWait({ prompt: "Say OK." }, 80_000).catch(() => undefined);
         } finally {
             await session.disconnect();
         }

nodejs/test/e2e/copilot_request_handler.e2e.test.ts

@@ -355,7 +355,11 @@ describe("CopilotRequestHandler — single subclass handles HTTP + WebSocket", a
         const session = await client.createSession({ onPermissionRequest: approveAll });
         let resultJson = "";
         try {
-            const result = await session.sendAndWait({ prompt: "Say OK." });
+            // Use an explicit idle timeout aligned with this test's 90s vitest
+            // budget; the sendAndWait default (60s) is lower than the declared
+            // budget, so under heavy parallel CI load it could fire before the
+            // end-to-end runtime spawn + WS handshake + inference completed.
+            const result = await session.sendAndWait({ prompt: "Say OK." }, 80_000);
             resultJson = JSON.stringify(result);
         } finally {
             await session.disconnect();

nodejs/test/e2e/copilot_request_session_id.e2e.test.ts

@@ -258,7 +258,10 @@ describe("CopilotRequestHandler threads the runtime session id (CAPI + BYOK)", a
         capiSessionId = session.sessionId;
         let resultJson = "";
         try {
-            const result = await session.sendAndWait({ prompt: "Say OK." });
+            // Explicit idle timeout matching this test's 90s vitest budget; the
+            // sendAndWait default (60s) is lower than the declared budget and
+            // can fire prematurely under heavy parallel CI load.
+            const result = await session.sendAndWait({ prompt: "Say OK." }, 80_000);
             resultJson = JSON.stringify(result);
         } finally {
             await session.disconnect();
@@ -298,7 +301,10 @@ describe("CopilotRequestHandler threads the runtime session id (CAPI + BYOK)", a
         const byokSessionId = session.sessionId;
         let resultJson = "";
         try {
-            const result = await session.sendAndWait({ prompt: "Say OK." });
+            // Explicit idle timeout matching this test's 90s vitest budget; the
+            // sendAndWait default (60s) is lower than the declared budget and
+            // can fire prematurely under heavy parallel CI load.
+            const result = await session.sendAndWait({ prompt: "Say OK." }, 80_000);
             resultJson = JSON.stringify(result);
         } finally {
             await session.disconnect();