From b7e86b5ba4f98a74644482c07109f612f3d49e29 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 17:04:41 -0400
Subject: [PATCH 1/9] Instrument Node in-process test stalls

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 .github/workflows/nodejs-sdk-tests.yml    |  1 +
 nodejs/src/client.ts                      | 32 +++++++++++++++++++++++
 nodejs/src/ffiRuntimeHost.ts              | 21 +++++++++++++++
 nodejs/src/session.ts                     | 16 ++++++++++++
 nodejs/test/e2e/harness/sdkTestContext.ts | 27 +++++++++++++++----
 5 files changed, 92 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/nodejs-sdk-tests.yml b/.github/workflows/nodejs-sdk-tests.yml
index 647345e0ea..f5a72a94d0 100644
--- a/.github/workflows/nodejs-sdk-tests.yml
+++ b/.github/workflows/nodejs-sdk-tests.yml
@@ -80,6 +80,7 @@ jobs:
         if: matrix.transport == 'inprocess'
         run: |
           echo "COPILOT_SDK_DEFAULT_CONNECTION=inprocess" >> "$GITHUB_ENV"
+          echo "COPILOT_SDK_TEST_DIAGNOSTICS=1" >> "$GITHUB_ENV"
 
       - name: Run Node.js SDK tests
         env:
diff --git a/nodejs/src/client.ts b/nodejs/src/client.ts
index 65784569cc..2c6b566fec 100644
--- a/nodejs/src/client.ts
+++ b/nodejs/src/client.ts
@@ -121,6 +121,14 @@ async function withTimeout<T>(promise: Promise<T>, timeoutMs: number, message: s
     }
 }
 
+function logTestDiagnostic(message: string, startMs?: number): void {
+    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
+        return;
+    }
+    const elapsed = startMs === undefined ? "" : ` elapsed=${Date.now() - startMs}ms`;
+    process.stderr.write(`[copilot-sdk-diagnostic pid=${process.pid}] ${message}${elapsed}\n`);
+}
+
 async function waitForChildExit(child: ChildProcess, timeoutMs: number): Promise<boolean> {
     if (child.exitCode != null || child.signalCode != null) {
         return true;
@@ -877,21 +885,26 @@ export class CopilotClient {
             return;
         }
 
+        const startMs = Date.now();
+        logTestDiagnostic(`client.start begin transport=${this.connectionConfig.kind}`);
         this.state = "connecting";
 
         try {
             // Only start CLI server process if not connecting to external server
             if (this.connectionConfig.kind === "inprocess") {
                 await this.startInProcessFfi();
+                logTestDiagnostic("client.start FFI host ready", startMs);
             } else if (!this.isExternalServer) {
                 await this.startCLIServer();
             }
 
             // Connect to the server
             await this.connectToServer();
+            logTestDiagnostic("client.start JSON-RPC connected", startMs);
 
             // Verify protocol version compatibility
             await this.verifyProtocolVersion();
+            logTestDiagnostic("client.start protocol verified", startMs);
 
             // If a session filesystem provider was configured, register it
             if (this.sessionFsConfig) {
@@ -911,8 +924,10 @@ export class CopilotClient {
             }
 
             this.state = "connected";
+            logTestDiagnostic("client.start complete", startMs);
         } catch (error) {
             this.state = "error";
+            logTestDiagnostic("client.start failed", startMs);
             throw error;
         }
     }
@@ -942,10 +957,12 @@ export class CopilotClient {
      * ```
      */
     async stop(): Promise<Error[]> {
+        const stopMs = Date.now();
         const errors: Error[] = [];
 
         // Disconnect all active sessions with retry logic
         const activeSessions = [...this.sessions.values()];
+        logTestDiagnostic(`client.stop begin sessions=${activeSessions.length}`);
         // TEMPORARY: over the in-process (FFI) transport the runtime shares this
         // process, so a turn still running when the runtime disposes the session
         // can leave that session's SQLite session.db handle open — it isn't
@@ -959,6 +976,7 @@ export class CopilotClient {
         // may still resume. Remove once the runtime cleans up fully on shutdown.
         if (this.connectionConfig.kind === "inprocess") {
             await Promise.allSettled(activeSessions.map((session) => session.abort()));
+            logTestDiagnostic("client.stop session aborts settled", stopMs);
         }
         for (const session of activeSessions) {
             const sessionId = session.sessionId;
@@ -989,6 +1007,7 @@ export class CopilotClient {
                 );
             }
         }
+        logTestDiagnostic("client.stop session disconnects complete", stopMs);
         for (const session of activeSessions) {
             session._markDisconnected();
         }
@@ -1023,6 +1042,7 @@ export class CopilotClient {
                 );
             }
         }
+        logTestDiagnostic("client.stop runtime shutdown stage complete", stopMs);
 
         // Close connection. Suppress writer failures first: tearing down the
         // transport can reject an in-flight server→client response write with
@@ -1046,6 +1066,7 @@ export class CopilotClient {
             this._rpc = null;
             this._internalRpc = null;
         }
+        logTestDiagnostic("client.stop JSON-RPC disposed", stopMs);
 
         // Clear models cache
         this.modelsCache = null;
@@ -1104,7 +1125,9 @@ export class CopilotClient {
             const host = this.ffiHost;
             this.ffiHost = null;
             try {
+                logTestDiagnostic("client.stop FFI host dispose begin", stopMs);
                 host.dispose();
+                logTestDiagnostic("client.stop FFI host dispose complete", stopMs);
             } catch (error) {
                 errors.push(
                     new Error(
@@ -1122,6 +1145,7 @@ export class CopilotClient {
         this.runtimePort = null;
         this.stderrBuffer = "";
         this.processExitPromise = null;
+        logTestDiagnostic(`client.stop complete errors=${errors.length}`, stopMs);
 
         return errors;
     }
@@ -1416,8 +1440,11 @@ export class CopilotClient {
     }
 
     async createSession(config: SessionConfig): Promise<CopilotSession> {
+        const createMs = Date.now();
+        logTestDiagnostic(`client.createSession begin connected=${this.connection !== null}`);
         if (!this.connection) {
             await this.start();
+            logTestDiagnostic("client.createSession client started", createMs);
         }
 
         config = { ...this.configDefaultsForMode(), ...config };
@@ -1588,6 +1615,7 @@ export class CopilotClient {
                 expAssignments: config.expAssignments,
                 enableManagedSettings: config.enableManagedSettings,
             });
+            logTestDiagnostic("client.createSession RPC complete", createMs);
 
             const {
                 sessionId: returnedSessionId,
@@ -1629,6 +1657,7 @@ export class CopilotClient {
             throw e;
         }
 
+        logTestDiagnostic("client.createSession complete", createMs);
         return session;
     }
 
@@ -2560,6 +2589,8 @@ export class CopilotClient {
 
     /** Starts the in-process FFI runtime with SDK-managed typed options. */
     private async startInProcessFfi(): Promise<void> {
+        const startMs = Date.now();
+        logTestDiagnostic("client.startInProcessFfi begin");
         const entrypoint = this.resolveCliPathForFfi();
         // Load the FFI host lazily so the native `koffi` addon (and its
         // platform-specific `koffi.node`) is only loaded on the in-process path;
@@ -2603,6 +2634,7 @@ export class CopilotClient {
         );
         this.ffiHost = host;
         await host.start();
+        logTestDiagnostic("client.startInProcessFfi complete", startMs);
     }
 
     /**
diff --git a/nodejs/src/ffiRuntimeHost.ts b/nodejs/src/ffiRuntimeHost.ts
index a92aa1589a..790e20f5b1 100644
--- a/nodejs/src/ffiRuntimeHost.ts
+++ b/nodejs/src/ffiRuntimeHost.ts
@@ -28,6 +28,14 @@ const SYMBOL_PREFIX = "copilot_runtime_";
 // connection is open (see start()); the exact interval is irrelevant.
 const KEEP_ALIVE_INTERVAL_MS = 1 << 30;
 
+function logTestDiagnostic(message: string, startMs?: number): void {
+    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
+        return;
+    }
+    const elapsed = startMs === undefined ? "" : ` elapsed=${Date.now() - startMs}ms`;
+    process.stderr.write(`[copilot-sdk-diagnostic pid=${process.pid}] ${message}${elapsed}\n`);
+}
+
 type KoffiFunction = ReturnType<ReturnType<typeof koffi.load>["func"]>;
 type KoffiType = ReturnType<typeof koffi.pointer>;
 type KoffiRegisteredCallback = ReturnType<typeof koffi.register>;
@@ -186,6 +194,8 @@ export class FfiRuntimeHost {
      * waits for readiness, and opens the FFI JSON-RPC connection.
      */
     async start(): Promise<void> {
+        const startMs = Date.now();
+        logTestDiagnostic("ffi.start begin");
         const argvJson = buildArgvJson(this.cliEntrypoint, this.args);
         const envJson = buildEnvJson(this.environment);
 
@@ -216,6 +226,7 @@ export class FfiRuntimeHost {
                 `copilot_runtime_host_start failed (library '${this.libraryPath}', entrypoint '${this.cliEntrypoint}').`
             );
         }
+        logTestDiagnostic("ffi.start hostStart complete", startMs);
 
         this.outboundCallback = koffi.register(
             (_userData: unknown, bytesPtr: unknown, bytesLen: number | bigint) =>
@@ -240,6 +251,7 @@ export class FfiRuntimeHost {
             this.serverId = 0;
             throw new Error("copilot_runtime_connection_open failed.");
         }
+        logTestDiagnostic("ffi.start connectionOpen complete", startMs);
 
         // The in-process transport has no socket/pipe handle to keep the Node event loop
         // alive while the SDK is idle awaiting a server→client frame. koffi delivers the
@@ -311,6 +323,10 @@ export class FfiRuntimeHost {
             return;
         }
         this.disposed = true;
+        const disposeMs = Date.now();
+        logTestDiagnostic(
+            `ffi.dispose begin server=${this.serverId !== 0} connection=${this.connectionId !== 0}`
+        );
 
         if (this.keepAliveTimer !== undefined) {
             clearInterval(this.keepAliveTimer);
@@ -319,8 +335,10 @@ export class FfiRuntimeHost {
 
         try {
             if (this.connectionId) {
+                logTestDiagnostic("ffi.dispose connectionClose begin", disposeMs);
                 this.lib.connectionClose(this.connectionId);
                 this.connectionId = 0;
+                logTestDiagnostic("ffi.dispose connectionClose complete", disposeMs);
             }
         } catch {
             // Ignore teardown failures.
@@ -328,8 +346,10 @@ export class FfiRuntimeHost {
 
         try {
             if (this.serverId) {
+                logTestDiagnostic("ffi.dispose hostShutdown begin", disposeMs);
                 this.lib.hostShutdown(this.serverId);
                 this.serverId = 0;
+                logTestDiagnostic("ffi.dispose hostShutdown complete", disposeMs);
             }
         } catch {
             // Ignore teardown failures.
@@ -337,5 +357,6 @@ export class FfiRuntimeHost {
 
         this.receiveStream.end();
         this.unregisterCallback();
+        logTestDiagnostic("ffi.dispose complete", disposeMs);
     }
 }
diff --git a/nodejs/src/session.ts b/nodejs/src/session.ts
index 1f71209de8..a6c7d8291a 100644
--- a/nodejs/src/session.ts
+++ b/nodejs/src/session.ts
@@ -61,6 +61,14 @@ import type {
     UserInputResponse,
 } from "./types.js";
 
+function logTestDiagnostic(message: string, startMs?: number): void {
+    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
+        return;
+    }
+    const elapsed = startMs === undefined ? "" : ` elapsed=${Date.now() - startMs}ms`;
+    process.stderr.write(`[copilot-sdk-diagnostic pid=${process.pid}] ${message}${elapsed}\n`);
+}
+
 /**
  * Convert a raw hook input received over the wire into its public-facing shape.
  * This deserializes the numeric Unix-ms `timestamp` field on BaseHookInput
@@ -292,6 +300,8 @@ export class CopilotSession {
         optionsOrPrompt: MessageOptions | string,
         timeout?: number
     ): Promise<AssistantMessageEvent | undefined> {
+        const sendMs = Date.now();
+        logTestDiagnostic(`session.sendAndWait begin session=${this.sessionId}`);
         const options: MessageOptions =
             typeof optionsOrPrompt === "string" ? { prompt: optionsOrPrompt } : optionsOrPrompt;
         const effectiveTimeout = timeout ?? 60_000;
@@ -322,6 +332,10 @@ export class CopilotSession {
         let timeoutId: ReturnType<typeof setTimeout> | undefined;
         try {
             await this.send(options);
+            logTestDiagnostic(
+                `session.sendAndWait send complete session=${this.sessionId}`,
+                sendMs
+            );
 
             const timeoutPromise = new Promise<never>((_, reject) => {
                 timeoutId = setTimeout(
@@ -335,6 +349,7 @@ export class CopilotSession {
                 );
             });
             await Promise.race([idlePromise, timeoutPromise]);
+            logTestDiagnostic(`session.sendAndWait idle session=${this.sessionId}`, sendMs);
 
             return lastAssistantMessage;
         } finally {
@@ -342,6 +357,7 @@ export class CopilotSession {
                 clearTimeout(timeoutId);
             }
             unsubscribe();
+            logTestDiagnostic(`session.sendAndWait complete session=${this.sessionId}`, sendMs);
         }
     }
 
diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index 624450e47c..12a7781bac 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -16,6 +16,21 @@ import { formatError, retry } from "./sdkTestHelper";
 export const isCI = process.env.GITHUB_ACTIONS === "true";
 export const DEFAULT_GITHUB_TOKEN = "fake-token-for-e2e-tests";
 
+async function runDiagnosticPhase<T>(name: string, action: () => Promise<T>): Promise<T> {
+    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
+        return action();
+    }
+    const startMs = Date.now();
+    console.error(`[sdk-test-diagnostic pid=${process.pid}] ${name} begin`);
+    try {
+        return await action();
+    } finally {
+        console.error(
+            `[sdk-test-diagnostic pid=${process.pid}] ${name} complete elapsed=${Date.now() - startMs}ms`
+        );
+    }
+}
+
 /**
  * True when the E2E suite is running over the in-process (FFI) transport
  * (COPILOT_SDK_DEFAULT_CONNECTION=inprocess). Use with `it.skipIf` / `describe.skipIf`
@@ -291,11 +306,13 @@ export async function createSdkTestContext({
     });
 
     afterAll(async () => {
-        await copilotClient.stop();
-        await openAiEndpoint.stop(anyTestFailed);
-        await rmDir("remove e2e test copilotHomeDir", copilotHomeDir);
-        await rmDir("remove e2e test homeDir", homeDir);
-        await rmDir("remove e2e test workDir", workDir);
+        await runDiagnosticPhase("client.stop", () => copilotClient.stop());
+        await runDiagnosticPhase("proxy.stop", () => openAiEndpoint.stop(anyTestFailed));
+        await runDiagnosticPhase("remove copilotHomeDir", () =>
+            rmDir("remove e2e test copilotHomeDir", copilotHomeDir)
+        );
+        await runDiagnosticPhase("remove homeDir", () => rmDir("remove e2e test homeDir", homeDir));
+        await runDiagnosticPhase("remove workDir", () => rmDir("remove e2e test workDir", workDir));
     });
 
     return harness;

From 2add01567130f4b617adb22717175fd4b5707b6c Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 17:30:15 -0400
Subject: [PATCH 2/9] Trace runtime and proxy during Node stalls

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 nodejs/test/e2e/harness/sdkTestContext.ts |  3 ++-
 test/harness/capturingHttpProxy.ts        | 17 +++++++++++++++++
 2 files changed, 19 insertions(+), 1 deletion(-)

diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index 12a7781bac..e2e5206b3f 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -230,7 +230,8 @@ export async function createSdkTestContext({
             // beforeEach below), so the worker inherits it; passing a per-client env here
             // would have no effect (and is rejected by the in-process transport).
             env: effectiveInProcess ? undefined : mergedEnv,
-            logLevel: logLevel || "error",
+            logLevel:
+                process.env.COPILOT_SDK_TEST_DIAGNOSTICS === "1" ? "debug" : (logLevel ?? "error"),
             connection: effectiveConnection,
             gitHubToken: authTokenToUse,
             ...rest,
diff --git a/test/harness/capturingHttpProxy.ts b/test/harness/capturingHttpProxy.ts
index edccca4ead..4a9f376524 100644
--- a/test/harness/capturingHttpProxy.ts
+++ b/test/harness/capturingHttpProxy.ts
@@ -32,6 +32,13 @@ export class CapturingHttpProxy {
       req.on("end", () => {
         const body = Buffer.concat(chunks).toString("utf8");
         const startTime = Date.now();
+        const diagnosticsEnabled =
+          process.env.COPILOT_SDK_TEST_DIAGNOSTICS === "1";
+        if (diagnosticsEnabled) {
+          console.error(
+            `[proxy-diagnostic pid=${process.pid}] request begin ${req.method || "GET"} ${req.url || "/"}`,
+          );
+        }
 
         const capturedRequest: CapturedRequest = {
           method: req.method || "GET",
@@ -86,6 +93,11 @@ export class CapturingHttpProxy {
             };
 
             exchange.durationMs = endTime - startTime;
+            if (diagnosticsEnabled) {
+              console.error(
+                `[proxy-diagnostic pid=${process.pid}] request complete ${capturedRequest.method} ${capturedRequest.url} status=${responseStatusCode || 500} elapsed=${exchange.durationMs}ms`,
+              );
+            }
 
             res.end();
           },
@@ -105,6 +117,11 @@ export class CapturingHttpProxy {
             };
 
             exchange.durationMs = endTime - startTime;
+            if (diagnosticsEnabled) {
+              console.error(
+                `[proxy-diagnostic pid=${process.pid}] request error ${capturedRequest.method} ${capturedRequest.url} elapsed=${exchange.durationMs}ms`,
+              );
+            }
 
             res.writeHead(exchange.response.statusCode, errorHeaders);
             res.end("Proxy error");

From 9f19df2a8dd9e759a57281286332000ebebbfb7b Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 17:59:45 -0400
Subject: [PATCH 3/9] Create focused Windows in-process stress run

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 .github/workflows/nodejs-sdk-tests.yml    | 11 ++++++++---
 nodejs/test/e2e/harness/sdkTestContext.ts |  3 +--
 2 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/nodejs-sdk-tests.yml b/.github/workflows/nodejs-sdk-tests.yml
index f5a72a94d0..496094de0a 100644
--- a/.github/workflows/nodejs-sdk-tests.yml
+++ b/.github/workflows/nodejs-sdk-tests.yml
@@ -38,8 +38,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-        transport: ["default", "inprocess"]
+        os: [windows-latest]
+        transport: ["inprocess"]
     runs-on: ${{ matrix.os }}
     defaults:
       run:
@@ -85,4 +85,9 @@ jobs:
       - name: Run Node.js SDK tests
         env:
           COPILOT_HMAC_KEY: ${{ secrets.COPILOT_DEVELOPER_CLI_INTEGRATION_HMAC_KEY }}
-        run: npm test
+        run: >-
+          npm test --
+          test/e2e/builtin_tools.e2e.test.ts
+          test/e2e/hooks.e2e.test.ts
+          test/e2e/agent_and_compact_rpc.e2e.test.ts
+          test/e2e/session.e2e.test.ts
diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index e2e5206b3f..9b05f3c5a8 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -230,8 +230,7 @@ export async function createSdkTestContext({
             // beforeEach below), so the worker inherits it; passing a per-client env here
             // would have no effect (and is rejected by the in-process transport).
             env: effectiveInProcess ? undefined : mergedEnv,
-            logLevel:
-                process.env.COPILOT_SDK_TEST_DIAGNOSTICS === "1" ? "debug" : (logLevel ?? "error"),
+            logLevel: logLevel ?? "error",
             connection: effectiveConnection,
             gitHubToken: authTokenToUse,
             ...rest,

From 57d13046587e3283c85aa74365409633fdd19cd4 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 18:04:50 -0400
Subject: [PATCH 4/9] Measure Windows test resource pressure

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

Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 .github/workflows/nodejs-sdk-tests.yml | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/nodejs-sdk-tests.yml b/.github/workflows/nodejs-sdk-tests.yml
index 496094de0a..1385fe2979 100644
--- a/.github/workflows/nodejs-sdk-tests.yml
+++ b/.github/workflows/nodejs-sdk-tests.yml
@@ -85,9 +85,16 @@ jobs:
       - name: Run Node.js SDK tests
         env:
           COPILOT_HMAC_KEY: ${{ secrets.COPILOT_DEVELOPER_CLI_INTEGRATION_HMAC_KEY }}
-        run: >-
-          npm test --
-          test/e2e/builtin_tools.e2e.test.ts
-          test/e2e/hooks.e2e.test.ts
-          test/e2e/agent_and_compact_rpc.e2e.test.ts
-          test/e2e/session.e2e.test.ts
+        run: |
+          pwsh.exe -NoProfile -Command '
+            while ($true) {
+              $cpu = [math]::Round((Get-Counter "\Processor(_Total)\% Processor Time").CounterSamples.CookedValue)
+              $memory = [math]::Round((Get-Counter "\Memory\Available MBytes").CounterSamples.CookedValue)
+              $nodeCount = @(Get-Process -Name node -ErrorAction SilentlyContinue).Count
+              Write-Host "[resource-diagnostic] $(Get-Date -Format o) cpu=$cpu availableMb=$memory nodeProcesses=$nodeCount"
+              Start-Sleep -Seconds 5
+            }
+          ' &
+          monitor_pid=$!
+          trap 'kill "$monitor_pid"' EXIT
+          npm test

From d4c997a82bca17d6ad2af862888cbe1a7c6a4b91 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 18:26:32 -0400
Subject: [PATCH 5/9] Preserve runtime diagnostics on failure

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

Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 .github/workflows/nodejs-sdk-tests.yml    | 43 ++++++++++++++++-------
 nodejs/test/e2e/harness/sdkTestContext.ts | 14 +++++---
 2 files changed, 40 insertions(+), 17 deletions(-)

diff --git a/.github/workflows/nodejs-sdk-tests.yml b/.github/workflows/nodejs-sdk-tests.yml
index 1385fe2979..410618d26d 100644
--- a/.github/workflows/nodejs-sdk-tests.yml
+++ b/.github/workflows/nodejs-sdk-tests.yml
@@ -10,19 +10,19 @@ on:
   pull_request:
     types: [opened, synchronize, reopened, ready_for_review]
     paths:
-      - 'nodejs/**'
-      - 'test/**'
-      - '.github/workflows/nodejs-sdk-tests.yml'
-      - '!nodejs/scripts/**'
-      - '!**/*.md'
-      - '!**/LICENSE*'
-      - '!**/.gitignore'
-      - '!**/.editorconfig'
-      - '!**/*.png'
-      - '!**/*.jpg'
-      - '!**/*.jpeg'
-      - '!**/*.gif'
-      - '!**/*.svg'
+      - "nodejs/**"
+      - "test/**"
+      - ".github/workflows/nodejs-sdk-tests.yml"
+      - "!nodejs/scripts/**"
+      - "!**/*.md"
+      - "!**/LICENSE*"
+      - "!**/.gitignore"
+      - "!**/.editorconfig"
+      - "!**/*.png"
+      - "!**/*.jpg"
+      - "!**/*.jpeg"
+      - "!**/*.gif"
+      - "!**/*.svg"
   workflow_dispatch:
   merge_group:
 
@@ -85,6 +85,7 @@ jobs:
       - name: Run Node.js SDK tests
         env:
           COPILOT_HMAC_KEY: ${{ secrets.COPILOT_DEVELOPER_CLI_INTEGRATION_HMAC_KEY }}
+          COPILOT_SDK_PRESERVE_TEST_HOME: "1"
         run: |
           pwsh.exe -NoProfile -Command '
             while ($true) {
@@ -98,3 +99,19 @@ jobs:
           monitor_pid=$!
           trap 'kill "$monitor_pid"' EXIT
           npm test
+
+      - name: Collect runtime diagnostics
+        if: failure()
+        shell: pwsh
+        run: |
+          $paths = Get-ChildItem $env:TEMP -Directory -Filter "copilot-test-home-*"
+          if ($paths) {
+            Compress-Archive -Path $paths.FullName -DestinationPath runtime-diagnostics.zip
+          }
+
+      - name: Upload runtime diagnostics
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: nodejs-runtime-diagnostics
+          path: nodejs/runtime-diagnostics.zip
diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index 9b05f3c5a8..c280ee903c 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -230,7 +230,9 @@ export async function createSdkTestContext({
             // beforeEach below), so the worker inherits it; passing a per-client env here
             // would have no effect (and is rejected by the in-process transport).
             env: effectiveInProcess ? undefined : mergedEnv,
-            logLevel: logLevel ?? "error",
+            logLevel:
+                logLevel ??
+                (process.env.COPILOT_SDK_PRESERVE_TEST_HOME === "1" ? "debug" : "error"),
             connection: effectiveConnection,
             gitHubToken: authTokenToUse,
             ...rest,
@@ -308,9 +310,13 @@ export async function createSdkTestContext({
     afterAll(async () => {
         await runDiagnosticPhase("client.stop", () => copilotClient.stop());
         await runDiagnosticPhase("proxy.stop", () => openAiEndpoint.stop(anyTestFailed));
-        await runDiagnosticPhase("remove copilotHomeDir", () =>
-            rmDir("remove e2e test copilotHomeDir", copilotHomeDir)
-        );
+        if (anyTestFailed && process.env.COPILOT_SDK_PRESERVE_TEST_HOME === "1") {
+            console.error(`[sdk-test-diagnostic pid=${process.pid}] preserved ${copilotHomeDir}`);
+        } else {
+            await runDiagnosticPhase("remove copilotHomeDir", () =>
+                rmDir("remove e2e test copilotHomeDir", copilotHomeDir)
+            );
+        }
         await runDiagnosticPhase("remove homeDir", () => rmDir("remove e2e test homeDir", homeDir));
         await runDiagnosticPhase("remove workDir", () => rmDir("remove e2e test workDir", workDir));
     });

From 977de35fb27116ce346611f883f5e7e025dde333 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 18:39:18 -0400
Subject: [PATCH 6/9] Trace Windows session database locks

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

Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 nodejs/test/e2e/harness/sdkTestContext.ts | 34 ++++++++++++++++++++++-
 1 file changed, 33 insertions(+), 1 deletion(-)

diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index c280ee903c..2e786182a7 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -6,6 +6,7 @@ import fs, { realpathSync } from "fs";
 import { rm } from "fs/promises";
 import os from "os";
 import { basename, dirname, join, resolve } from "path";
+import { execFileSync } from "node:child_process";
 import { rimraf } from "rimraf";
 import { fileURLToPath } from "url";
 import { afterAll, afterEach, beforeEach, onTestFailed, TestContext } from "vitest";
@@ -349,8 +350,39 @@ async function rmDir(message: string, path: string): Promise<void> {
     // be removed (e.g. CLI background writer racing with cleanup), warn and
     // continue rather than failing the whole test run — the OS / CI runner
     // will reclaim the temp dir on shutdown.
+    let reportedLock = false;
     try {
-        await retry(message, () => rm(path, { recursive: true, force: true }), 30, 1000);
+        await retry(
+            message,
+            async () => {
+                try {
+                    await rm(path, { recursive: true, force: true });
+                } catch (error) {
+                    if (
+                        !reportedLock &&
+                        process.platform === "win32" &&
+                        process.env.COPILOT_SDK_TEST_DIAGNOSTICS === "1"
+                    ) {
+                        reportedLock = true;
+                        const processes = execFileSync(
+                            "powershell.exe",
+                            [
+                                "-NoProfile",
+                                "-Command",
+                                "Get-CimInstance Win32_Process | Where-Object { $_.Name -in @('node.exe', 'cmd.exe', 'copilot.exe') } | Select-Object ProcessId, ParentProcessId, CreationDate, Name, CommandLine | ConvertTo-Json -Compress",
+                            ],
+                            { encoding: "utf8" }
+                        );
+                        console.error(
+                            `[sdk-test-diagnostic pid=${process.pid}] remove failed path=${path} error=${formatError(error)} processes=${processes.trim()}`
+                        );
+                    }
+                    throw error;
+                }
+            },
+            30,
+            1000
+        );
     } catch (error) {
         console.warn(
             `WARN: ${message} failed; leaving temp dir for OS cleanup: ${formatError(error)}`

From a4f17fbdc622756230ac21f07541daf6ea6d05b8 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 19:05:46 -0400
Subject: [PATCH 7/9] Avoid perturbing runtime lock timing

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

Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 nodejs/test/e2e/harness/sdkTestContext.ts | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index 2e786182a7..6b837eb8fd 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -231,9 +231,7 @@ export async function createSdkTestContext({
             // beforeEach below), so the worker inherits it; passing a per-client env here
             // would have no effect (and is rejected by the in-process transport).
             env: effectiveInProcess ? undefined : mergedEnv,
-            logLevel:
-                logLevel ??
-                (process.env.COPILOT_SDK_PRESERVE_TEST_HOME === "1" ? "debug" : "error"),
+            logLevel: logLevel ?? "error",
             connection: effectiveConnection,
             gitHubToken: authTokenToUse,
             ...rest,

From 5e60dadac2ee55f0810be1f2ad04385ffa38ef07 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 19:26:37 -0400
Subject: [PATCH 8/9] Avoid Windows in-process teardown deadlock

Do not retry removal of the in-process runtime's session home while its Vitest worker still owns a locked session database. Retrying until the hook timeout prevents the worker from exiting and releasing the lock.

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

Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 .github/workflows/nodejs-sdk-tests.yml    | 62 +++++--------------
 nodejs/src/client.ts                      | 32 ----------
 nodejs/src/ffiRuntimeHost.ts              | 21 -------
 nodejs/src/session.ts                     | 16 -----
 nodejs/test/e2e/harness/sdkTestContext.ts | 75 +++++------------------
 test/harness/capturingHttpProxy.ts        | 17 -----
 6 files changed, 30 insertions(+), 193 deletions(-)

diff --git a/.github/workflows/nodejs-sdk-tests.yml b/.github/workflows/nodejs-sdk-tests.yml
index 410618d26d..647345e0ea 100644
--- a/.github/workflows/nodejs-sdk-tests.yml
+++ b/.github/workflows/nodejs-sdk-tests.yml
@@ -10,19 +10,19 @@ on:
   pull_request:
     types: [opened, synchronize, reopened, ready_for_review]
     paths:
-      - "nodejs/**"
-      - "test/**"
-      - ".github/workflows/nodejs-sdk-tests.yml"
-      - "!nodejs/scripts/**"
-      - "!**/*.md"
-      - "!**/LICENSE*"
-      - "!**/.gitignore"
-      - "!**/.editorconfig"
-      - "!**/*.png"
-      - "!**/*.jpg"
-      - "!**/*.jpeg"
-      - "!**/*.gif"
-      - "!**/*.svg"
+      - 'nodejs/**'
+      - 'test/**'
+      - '.github/workflows/nodejs-sdk-tests.yml'
+      - '!nodejs/scripts/**'
+      - '!**/*.md'
+      - '!**/LICENSE*'
+      - '!**/.gitignore'
+      - '!**/.editorconfig'
+      - '!**/*.png'
+      - '!**/*.jpg'
+      - '!**/*.jpeg'
+      - '!**/*.gif'
+      - '!**/*.svg'
   workflow_dispatch:
   merge_group:
 
@@ -38,8 +38,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [windows-latest]
-        transport: ["inprocess"]
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        transport: ["default", "inprocess"]
     runs-on: ${{ matrix.os }}
     defaults:
       run:
@@ -80,38 +80,8 @@ jobs:
         if: matrix.transport == 'inprocess'
         run: |
           echo "COPILOT_SDK_DEFAULT_CONNECTION=inprocess" >> "$GITHUB_ENV"
-          echo "COPILOT_SDK_TEST_DIAGNOSTICS=1" >> "$GITHUB_ENV"
 
       - name: Run Node.js SDK tests
         env:
           COPILOT_HMAC_KEY: ${{ secrets.COPILOT_DEVELOPER_CLI_INTEGRATION_HMAC_KEY }}
-          COPILOT_SDK_PRESERVE_TEST_HOME: "1"
-        run: |
-          pwsh.exe -NoProfile -Command '
-            while ($true) {
-              $cpu = [math]::Round((Get-Counter "\Processor(_Total)\% Processor Time").CounterSamples.CookedValue)
-              $memory = [math]::Round((Get-Counter "\Memory\Available MBytes").CounterSamples.CookedValue)
-              $nodeCount = @(Get-Process -Name node -ErrorAction SilentlyContinue).Count
-              Write-Host "[resource-diagnostic] $(Get-Date -Format o) cpu=$cpu availableMb=$memory nodeProcesses=$nodeCount"
-              Start-Sleep -Seconds 5
-            }
-          ' &
-          monitor_pid=$!
-          trap 'kill "$monitor_pid"' EXIT
-          npm test
-
-      - name: Collect runtime diagnostics
-        if: failure()
-        shell: pwsh
-        run: |
-          $paths = Get-ChildItem $env:TEMP -Directory -Filter "copilot-test-home-*"
-          if ($paths) {
-            Compress-Archive -Path $paths.FullName -DestinationPath runtime-diagnostics.zip
-          }
-
-      - name: Upload runtime diagnostics
-        if: failure()
-        uses: actions/upload-artifact@v4
-        with:
-          name: nodejs-runtime-diagnostics
-          path: nodejs/runtime-diagnostics.zip
+        run: npm test
diff --git a/nodejs/src/client.ts b/nodejs/src/client.ts
index 2c6b566fec..65784569cc 100644
--- a/nodejs/src/client.ts
+++ b/nodejs/src/client.ts
@@ -121,14 +121,6 @@ async function withTimeout<T>(promise: Promise<T>, timeoutMs: number, message: s
     }
 }
 
-function logTestDiagnostic(message: string, startMs?: number): void {
-    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
-        return;
-    }
-    const elapsed = startMs === undefined ? "" : ` elapsed=${Date.now() - startMs}ms`;
-    process.stderr.write(`[copilot-sdk-diagnostic pid=${process.pid}] ${message}${elapsed}\n`);
-}
-
 async function waitForChildExit(child: ChildProcess, timeoutMs: number): Promise<boolean> {
     if (child.exitCode != null || child.signalCode != null) {
         return true;
@@ -885,26 +877,21 @@ export class CopilotClient {
             return;
         }
 
-        const startMs = Date.now();
-        logTestDiagnostic(`client.start begin transport=${this.connectionConfig.kind}`);
         this.state = "connecting";
 
         try {
             // Only start CLI server process if not connecting to external server
             if (this.connectionConfig.kind === "inprocess") {
                 await this.startInProcessFfi();
-                logTestDiagnostic("client.start FFI host ready", startMs);
             } else if (!this.isExternalServer) {
                 await this.startCLIServer();
             }
 
             // Connect to the server
             await this.connectToServer();
-            logTestDiagnostic("client.start JSON-RPC connected", startMs);
 
             // Verify protocol version compatibility
             await this.verifyProtocolVersion();
-            logTestDiagnostic("client.start protocol verified", startMs);
 
             // If a session filesystem provider was configured, register it
             if (this.sessionFsConfig) {
@@ -924,10 +911,8 @@ export class CopilotClient {
             }
 
             this.state = "connected";
-            logTestDiagnostic("client.start complete", startMs);
         } catch (error) {
             this.state = "error";
-            logTestDiagnostic("client.start failed", startMs);
             throw error;
         }
     }
@@ -957,12 +942,10 @@ export class CopilotClient {
      * ```
      */
     async stop(): Promise<Error[]> {
-        const stopMs = Date.now();
         const errors: Error[] = [];
 
         // Disconnect all active sessions with retry logic
         const activeSessions = [...this.sessions.values()];
-        logTestDiagnostic(`client.stop begin sessions=${activeSessions.length}`);
         // TEMPORARY: over the in-process (FFI) transport the runtime shares this
         // process, so a turn still running when the runtime disposes the session
         // can leave that session's SQLite session.db handle open — it isn't
@@ -976,7 +959,6 @@ export class CopilotClient {
         // may still resume. Remove once the runtime cleans up fully on shutdown.
         if (this.connectionConfig.kind === "inprocess") {
             await Promise.allSettled(activeSessions.map((session) => session.abort()));
-            logTestDiagnostic("client.stop session aborts settled", stopMs);
         }
         for (const session of activeSessions) {
             const sessionId = session.sessionId;
@@ -1007,7 +989,6 @@ export class CopilotClient {
                 );
             }
         }
-        logTestDiagnostic("client.stop session disconnects complete", stopMs);
         for (const session of activeSessions) {
             session._markDisconnected();
         }
@@ -1042,7 +1023,6 @@ export class CopilotClient {
                 );
             }
         }
-        logTestDiagnostic("client.stop runtime shutdown stage complete", stopMs);
 
         // Close connection. Suppress writer failures first: tearing down the
         // transport can reject an in-flight server→client response write with
@@ -1066,7 +1046,6 @@ export class CopilotClient {
             this._rpc = null;
             this._internalRpc = null;
         }
-        logTestDiagnostic("client.stop JSON-RPC disposed", stopMs);
 
         // Clear models cache
         this.modelsCache = null;
@@ -1125,9 +1104,7 @@ export class CopilotClient {
             const host = this.ffiHost;
             this.ffiHost = null;
             try {
-                logTestDiagnostic("client.stop FFI host dispose begin", stopMs);
                 host.dispose();
-                logTestDiagnostic("client.stop FFI host dispose complete", stopMs);
             } catch (error) {
                 errors.push(
                     new Error(
@@ -1145,7 +1122,6 @@ export class CopilotClient {
         this.runtimePort = null;
         this.stderrBuffer = "";
         this.processExitPromise = null;
-        logTestDiagnostic(`client.stop complete errors=${errors.length}`, stopMs);
 
         return errors;
     }
@@ -1440,11 +1416,8 @@ export class CopilotClient {
     }
 
     async createSession(config: SessionConfig): Promise<CopilotSession> {
-        const createMs = Date.now();
-        logTestDiagnostic(`client.createSession begin connected=${this.connection !== null}`);
         if (!this.connection) {
             await this.start();
-            logTestDiagnostic("client.createSession client started", createMs);
         }
 
         config = { ...this.configDefaultsForMode(), ...config };
@@ -1615,7 +1588,6 @@ export class CopilotClient {
                 expAssignments: config.expAssignments,
                 enableManagedSettings: config.enableManagedSettings,
             });
-            logTestDiagnostic("client.createSession RPC complete", createMs);
 
             const {
                 sessionId: returnedSessionId,
@@ -1657,7 +1629,6 @@ export class CopilotClient {
             throw e;
         }
 
-        logTestDiagnostic("client.createSession complete", createMs);
         return session;
     }
 
@@ -2589,8 +2560,6 @@ export class CopilotClient {
 
     /** Starts the in-process FFI runtime with SDK-managed typed options. */
     private async startInProcessFfi(): Promise<void> {
-        const startMs = Date.now();
-        logTestDiagnostic("client.startInProcessFfi begin");
         const entrypoint = this.resolveCliPathForFfi();
         // Load the FFI host lazily so the native `koffi` addon (and its
         // platform-specific `koffi.node`) is only loaded on the in-process path;
@@ -2634,7 +2603,6 @@ export class CopilotClient {
         );
         this.ffiHost = host;
         await host.start();
-        logTestDiagnostic("client.startInProcessFfi complete", startMs);
     }
 
     /**
diff --git a/nodejs/src/ffiRuntimeHost.ts b/nodejs/src/ffiRuntimeHost.ts
index 790e20f5b1..a92aa1589a 100644
--- a/nodejs/src/ffiRuntimeHost.ts
+++ b/nodejs/src/ffiRuntimeHost.ts
@@ -28,14 +28,6 @@ const SYMBOL_PREFIX = "copilot_runtime_";
 // connection is open (see start()); the exact interval is irrelevant.
 const KEEP_ALIVE_INTERVAL_MS = 1 << 30;
 
-function logTestDiagnostic(message: string, startMs?: number): void {
-    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
-        return;
-    }
-    const elapsed = startMs === undefined ? "" : ` elapsed=${Date.now() - startMs}ms`;
-    process.stderr.write(`[copilot-sdk-diagnostic pid=${process.pid}] ${message}${elapsed}\n`);
-}
-
 type KoffiFunction = ReturnType<ReturnType<typeof koffi.load>["func"]>;
 type KoffiType = ReturnType<typeof koffi.pointer>;
 type KoffiRegisteredCallback = ReturnType<typeof koffi.register>;
@@ -194,8 +186,6 @@ export class FfiRuntimeHost {
      * waits for readiness, and opens the FFI JSON-RPC connection.
      */
     async start(): Promise<void> {
-        const startMs = Date.now();
-        logTestDiagnostic("ffi.start begin");
         const argvJson = buildArgvJson(this.cliEntrypoint, this.args);
         const envJson = buildEnvJson(this.environment);
 
@@ -226,7 +216,6 @@ export class FfiRuntimeHost {
                 `copilot_runtime_host_start failed (library '${this.libraryPath}', entrypoint '${this.cliEntrypoint}').`
             );
         }
-        logTestDiagnostic("ffi.start hostStart complete", startMs);
 
         this.outboundCallback = koffi.register(
             (_userData: unknown, bytesPtr: unknown, bytesLen: number | bigint) =>
@@ -251,7 +240,6 @@ export class FfiRuntimeHost {
             this.serverId = 0;
             throw new Error("copilot_runtime_connection_open failed.");
         }
-        logTestDiagnostic("ffi.start connectionOpen complete", startMs);
 
         // The in-process transport has no socket/pipe handle to keep the Node event loop
         // alive while the SDK is idle awaiting a server→client frame. koffi delivers the
@@ -323,10 +311,6 @@ export class FfiRuntimeHost {
             return;
         }
         this.disposed = true;
-        const disposeMs = Date.now();
-        logTestDiagnostic(
-            `ffi.dispose begin server=${this.serverId !== 0} connection=${this.connectionId !== 0}`
-        );
 
         if (this.keepAliveTimer !== undefined) {
             clearInterval(this.keepAliveTimer);
@@ -335,10 +319,8 @@ export class FfiRuntimeHost {
 
         try {
             if (this.connectionId) {
-                logTestDiagnostic("ffi.dispose connectionClose begin", disposeMs);
                 this.lib.connectionClose(this.connectionId);
                 this.connectionId = 0;
-                logTestDiagnostic("ffi.dispose connectionClose complete", disposeMs);
             }
         } catch {
             // Ignore teardown failures.
@@ -346,10 +328,8 @@ export class FfiRuntimeHost {
 
         try {
             if (this.serverId) {
-                logTestDiagnostic("ffi.dispose hostShutdown begin", disposeMs);
                 this.lib.hostShutdown(this.serverId);
                 this.serverId = 0;
-                logTestDiagnostic("ffi.dispose hostShutdown complete", disposeMs);
             }
         } catch {
             // Ignore teardown failures.
@@ -357,6 +337,5 @@ export class FfiRuntimeHost {
 
         this.receiveStream.end();
         this.unregisterCallback();
-        logTestDiagnostic("ffi.dispose complete", disposeMs);
     }
 }
diff --git a/nodejs/src/session.ts b/nodejs/src/session.ts
index a6c7d8291a..1f71209de8 100644
--- a/nodejs/src/session.ts
+++ b/nodejs/src/session.ts
@@ -61,14 +61,6 @@ import type {
     UserInputResponse,
 } from "./types.js";
 
-function logTestDiagnostic(message: string, startMs?: number): void {
-    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
-        return;
-    }
-    const elapsed = startMs === undefined ? "" : ` elapsed=${Date.now() - startMs}ms`;
-    process.stderr.write(`[copilot-sdk-diagnostic pid=${process.pid}] ${message}${elapsed}\n`);
-}
-
 /**
  * Convert a raw hook input received over the wire into its public-facing shape.
  * This deserializes the numeric Unix-ms `timestamp` field on BaseHookInput
@@ -300,8 +292,6 @@ export class CopilotSession {
         optionsOrPrompt: MessageOptions | string,
         timeout?: number
     ): Promise<AssistantMessageEvent | undefined> {
-        const sendMs = Date.now();
-        logTestDiagnostic(`session.sendAndWait begin session=${this.sessionId}`);
         const options: MessageOptions =
             typeof optionsOrPrompt === "string" ? { prompt: optionsOrPrompt } : optionsOrPrompt;
         const effectiveTimeout = timeout ?? 60_000;
@@ -332,10 +322,6 @@ export class CopilotSession {
         let timeoutId: ReturnType<typeof setTimeout> | undefined;
         try {
             await this.send(options);
-            logTestDiagnostic(
-                `session.sendAndWait send complete session=${this.sessionId}`,
-                sendMs
-            );
 
             const timeoutPromise = new Promise<never>((_, reject) => {
                 timeoutId = setTimeout(
@@ -349,7 +335,6 @@ export class CopilotSession {
                 );
             });
             await Promise.race([idlePromise, timeoutPromise]);
-            logTestDiagnostic(`session.sendAndWait idle session=${this.sessionId}`, sendMs);
 
             return lastAssistantMessage;
         } finally {
@@ -357,7 +342,6 @@ export class CopilotSession {
                 clearTimeout(timeoutId);
             }
             unsubscribe();
-            logTestDiagnostic(`session.sendAndWait complete session=${this.sessionId}`, sendMs);
         }
     }
 
diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index 6b837eb8fd..f2dd607edb 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -6,7 +6,6 @@ import fs, { realpathSync } from "fs";
 import { rm } from "fs/promises";
 import os from "os";
 import { basename, dirname, join, resolve } from "path";
-import { execFileSync } from "node:child_process";
 import { rimraf } from "rimraf";
 import { fileURLToPath } from "url";
 import { afterAll, afterEach, beforeEach, onTestFailed, TestContext } from "vitest";
@@ -17,21 +16,6 @@ import { formatError, retry } from "./sdkTestHelper";
 export const isCI = process.env.GITHUB_ACTIONS === "true";
 export const DEFAULT_GITHUB_TOKEN = "fake-token-for-e2e-tests";
 
-async function runDiagnosticPhase<T>(name: string, action: () => Promise<T>): Promise<T> {
-    if (process.env.COPILOT_SDK_TEST_DIAGNOSTICS !== "1") {
-        return action();
-    }
-    const startMs = Date.now();
-    console.error(`[sdk-test-diagnostic pid=${process.pid}] ${name} begin`);
-    try {
-        return await action();
-    } finally {
-        console.error(
-            `[sdk-test-diagnostic pid=${process.pid}] ${name} complete elapsed=${Date.now() - startMs}ms`
-        );
-    }
-}
-
 /**
  * True when the E2E suite is running over the in-process (FFI) transport
  * (COPILOT_SDK_DEFAULT_CONNECTION=inprocess). Use with `it.skipIf` / `describe.skipIf`
@@ -231,7 +215,7 @@ export async function createSdkTestContext({
             // beforeEach below), so the worker inherits it; passing a per-client env here
             // would have no effect (and is rejected by the in-process transport).
             env: effectiveInProcess ? undefined : mergedEnv,
-            logLevel: logLevel ?? "error",
+            logLevel: logLevel || "error",
             connection: effectiveConnection,
             gitHubToken: authTokenToUse,
             ...rest,
@@ -307,17 +291,17 @@ export async function createSdkTestContext({
     });
 
     afterAll(async () => {
-        await runDiagnosticPhase("client.stop", () => copilotClient.stop());
-        await runDiagnosticPhase("proxy.stop", () => openAiEndpoint.stop(anyTestFailed));
-        if (anyTestFailed && process.env.COPILOT_SDK_PRESERVE_TEST_HOME === "1") {
-            console.error(`[sdk-test-diagnostic pid=${process.pid}] preserved ${copilotHomeDir}`);
-        } else {
-            await runDiagnosticPhase("remove copilotHomeDir", () =>
-                rmDir("remove e2e test copilotHomeDir", copilotHomeDir)
-            );
-        }
-        await runDiagnosticPhase("remove homeDir", () => rmDir("remove e2e test homeDir", homeDir));
-        await runDiagnosticPhase("remove workDir", () => rmDir("remove e2e test workDir", workDir));
+        await copilotClient.stop();
+        await openAiEndpoint.stop(anyTestFailed);
+        // On Windows, the in-process runtime can keep session.db locked until this
+        // Vitest worker exits. Retrying here prevents that exit and deadlocks teardown.
+        await rmDir(
+            "remove e2e test copilotHomeDir",
+            copilotHomeDir,
+            isInProcess && process.platform === "win32" ? 1 : 30
+        );
+        await rmDir("remove e2e test homeDir", homeDir);
+        await rmDir("remove e2e test workDir", workDir);
     });
 
     return harness;
@@ -342,45 +326,14 @@ function getTrafficCapturePath(testContext: TestContext): string {
     return join(SNAPSHOTS_DIR, testFileName, `${taskNameAsFilename}.yaml`);
 }
 
-async function rmDir(message: string, path: string): Promise<void> {
+async function rmDir(message: string, path: string, maxTries = 30): Promise<void> {
     // Use longer retries to tolerate Windows holding SQLite session-store.db
     // open briefly after the CLI subprocess exits. If the temp dir still can't
     // be removed (e.g. CLI background writer racing with cleanup), warn and
     // continue rather than failing the whole test run — the OS / CI runner
     // will reclaim the temp dir on shutdown.
-    let reportedLock = false;
     try {
-        await retry(
-            message,
-            async () => {
-                try {
-                    await rm(path, { recursive: true, force: true });
-                } catch (error) {
-                    if (
-                        !reportedLock &&
-                        process.platform === "win32" &&
-                        process.env.COPILOT_SDK_TEST_DIAGNOSTICS === "1"
-                    ) {
-                        reportedLock = true;
-                        const processes = execFileSync(
-                            "powershell.exe",
-                            [
-                                "-NoProfile",
-                                "-Command",
-                                "Get-CimInstance Win32_Process | Where-Object { $_.Name -in @('node.exe', 'cmd.exe', 'copilot.exe') } | Select-Object ProcessId, ParentProcessId, CreationDate, Name, CommandLine | ConvertTo-Json -Compress",
-                            ],
-                            { encoding: "utf8" }
-                        );
-                        console.error(
-                            `[sdk-test-diagnostic pid=${process.pid}] remove failed path=${path} error=${formatError(error)} processes=${processes.trim()}`
-                        );
-                    }
-                    throw error;
-                }
-            },
-            30,
-            1000
-        );
+        await retry(message, () => rm(path, { recursive: true, force: true }), maxTries, 1000);
     } catch (error) {
         console.warn(
             `WARN: ${message} failed; leaving temp dir for OS cleanup: ${formatError(error)}`
diff --git a/test/harness/capturingHttpProxy.ts b/test/harness/capturingHttpProxy.ts
index 4a9f376524..edccca4ead 100644
--- a/test/harness/capturingHttpProxy.ts
+++ b/test/harness/capturingHttpProxy.ts
@@ -32,13 +32,6 @@ export class CapturingHttpProxy {
       req.on("end", () => {
         const body = Buffer.concat(chunks).toString("utf8");
         const startTime = Date.now();
-        const diagnosticsEnabled =
-          process.env.COPILOT_SDK_TEST_DIAGNOSTICS === "1";
-        if (diagnosticsEnabled) {
-          console.error(
-            `[proxy-diagnostic pid=${process.pid}] request begin ${req.method || "GET"} ${req.url || "/"}`,
-          );
-        }
 
         const capturedRequest: CapturedRequest = {
           method: req.method || "GET",
@@ -93,11 +86,6 @@ export class CapturingHttpProxy {
             };
 
             exchange.durationMs = endTime - startTime;
-            if (diagnosticsEnabled) {
-              console.error(
-                `[proxy-diagnostic pid=${process.pid}] request complete ${capturedRequest.method} ${capturedRequest.url} status=${responseStatusCode || 500} elapsed=${exchange.durationMs}ms`,
-              );
-            }
 
             res.end();
           },
@@ -117,11 +105,6 @@ export class CapturingHttpProxy {
             };
 
             exchange.durationMs = endTime - startTime;
-            if (diagnosticsEnabled) {
-              console.error(
-                `[proxy-diagnostic pid=${process.pid}] request error ${capturedRequest.method} ${capturedRequest.url} elapsed=${exchange.durationMs}ms`,
-              );
-            }
 
             res.writeHead(exchange.response.statusCode, errorHeaders);
             res.end("Proxy error");

From 9dc2c289e7867c414ed63ac7f9397090ae4364ac Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Wed, 15 Jul 2026 20:57:12 -0400
Subject: [PATCH 9/9] Clarify Windows teardown deadlock

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

Copilot-Session: 6e902b9a-5527-4a32-a5a3-e0bf5bfef3f7
---
 nodejs/test/e2e/harness/sdkTestContext.ts | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/nodejs/test/e2e/harness/sdkTestContext.ts b/nodejs/test/e2e/harness/sdkTestContext.ts
index f2dd607edb..bf62db4826 100644
--- a/nodejs/test/e2e/harness/sdkTestContext.ts
+++ b/nodejs/test/e2e/harness/sdkTestContext.ts
@@ -293,8 +293,10 @@ export async function createSdkTestContext({
     afterAll(async () => {
         await copilotClient.stop();
         await openAiEndpoint.stop(anyTestFailed);
-        // On Windows, the in-process runtime can keep session.db locked until this
-        // Vitest worker exits. Retrying here prevents that exit and deadlocks teardown.
+        // On Windows, this Vitest worker can retain the in-process runtime's session.db
+        // lock until the worker exits. Retrying from its afterAll hook cannot succeed:
+        // the hook waits for the lock, while the lock cannot clear until the hook returns
+        // and lets the worker exit.
         await rmDir(
             "remove e2e test copilotHomeDir",
             copilotHomeDir,