fix: prevent timer leaks and event-loop stalls in SSH timeout handling (#2200)

- Unref the SIGKILL timer in killWithTimeout() so it doesn't keep the
  event loop alive for 5 extra seconds after a timed-out process exits
- Wrap all setTimeout/clearTimeout pairs in try/finally across 6 cloud
  providers (12 call sites) to guarantee cleanup on exceptions
- Add missing 60s timeout guard to runSpriteSilent() which could hang
  indefinitely on unresponsive sprite processes

Agent: code-health

Co-authored-by: B <6723574+louisgv@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

packages/cli/src/aws/aws.ts

@@ -1047,10 +1047,13 @@ export async function runServer(cmd: string, timeoutSecs?: number): Promise<void
   );
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-  if (exitCode !== 0) {
-    throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+  try {
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+    }
+  } finally {
+    clearTimeout(timer);
   }
 }
 
@@ -1075,17 +1078,20 @@ export async function runServerCapture(cmd: string, timeoutSecs?: number): Promi
   );
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
-  const [stdout] = await Promise.all([
-    new Response(proc.stdout).text(),
-    new Response(proc.stderr).text(),
-  ]);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-  if (exitCode !== 0) {
-    throw new Error(`run_server_capture failed (exit ${exitCode})`);
+  try {
+    // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
+    const [stdout] = await Promise.all([
+      new Response(proc.stdout).text(),
+      new Response(proc.stderr).text(),
+    ]);
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server_capture failed (exit ${exitCode})`);
+    }
+    return stdout.trim();
+  } finally {
+    clearTimeout(timer);
   }
-  return stdout.trim();
 }
 
 export async function uploadFile(localPath: string, remotePath: string): Promise<void> {

packages/cli/src/daytona/daytona.ts

@@ -389,14 +389,15 @@ export async function runServer(cmd: string, timeoutSecs?: number): Promise<void
   }
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-
-  // Brief sleep to let gateway release connection slot
-  await sleep(1000);
-
-  if (exitCode !== 0) {
-    throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+  try {
+    const exitCode = await proc.exited;
+    // Brief sleep to let gateway release connection slot
+    await sleep(1000);
+    if (exitCode !== 0) {
+      throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+    }
+  } finally {
+    clearTimeout(timer);
   }
 }
 
@@ -426,20 +427,21 @@ export async function runServerCapture(cmd: string, timeoutSecs?: number): Promi
   }
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
-  const [stdout] = await Promise.all([
-    new Response(proc.stdout).text(),
-    new Response(proc.stderr).text(),
-  ]);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-
-  await sleep(1000);
-
-  if (exitCode !== 0) {
-    throw new Error(`run_server_capture failed (exit ${exitCode})`);
+  try {
+    // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
+    const [stdout] = await Promise.all([
+      new Response(proc.stdout).text(),
+      new Response(proc.stderr).text(),
+    ]);
+    const exitCode = await proc.exited;
+    await sleep(1000);
+    if (exitCode !== 0) {
+      throw new Error(`run_server_capture failed (exit ${exitCode})`);
+    }
+    return stdout.trim();
+  } finally {
+    clearTimeout(timer);
   }
-  return stdout.trim();
 }
 
 /**

packages/cli/src/digitalocean/digitalocean.ts

@@ -984,11 +984,13 @@ export async function runServer(cmd: string, timeoutSecs?: number, ip?: string):
 
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-
-  if (exitCode !== 0) {
-    throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+  try {
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+    }
+  } finally {
+    clearTimeout(timer);
   }
 }
 
@@ -1016,18 +1018,20 @@ export async function runServerCapture(cmd: string, timeoutSecs?: number, ip?: s
 
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
-  const [stdout] = await Promise.all([
-    new Response(proc.stdout).text(),
-    new Response(proc.stderr).text(),
-  ]);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-
-  if (exitCode !== 0) {
-    throw new Error(`run_server_capture failed (exit ${exitCode})`);
+  try {
+    // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
+    const [stdout] = await Promise.all([
+      new Response(proc.stdout).text(),
+      new Response(proc.stderr).text(),
+    ]);
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server_capture failed (exit ${exitCode})`);
+    }
+    return stdout.trim();
+  } finally {
+    clearTimeout(timer);
   }
-  return stdout.trim();
 }
 
 export async function uploadFile(localPath: string, remotePath: string, ip?: string): Promise<void> {

packages/cli/src/gcp/gcp.ts

@@ -848,10 +848,13 @@ export async function runServer(cmd: string, timeoutSecs?: number): Promise<void
   );
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-  if (exitCode !== 0) {
-    throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+  try {
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+    }
+  } finally {
+    clearTimeout(timer);
   }
 }
 
@@ -879,17 +882,20 @@ export async function runServerCapture(cmd: string, timeoutSecs?: number): Promi
   );
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
-  const [stdout] = await Promise.all([
-    new Response(proc.stdout).text(),
-    new Response(proc.stderr).text(),
-  ]);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-  if (exitCode !== 0) {
-    throw new Error(`run_server_capture failed (exit ${exitCode})`);
+  try {
+    // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
+    const [stdout] = await Promise.all([
+      new Response(proc.stdout).text(),
+      new Response(proc.stderr).text(),
+    ]);
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server_capture failed (exit ${exitCode})`);
+    }
+    return stdout.trim();
+  } finally {
+    clearTimeout(timer);
   }
-  return stdout.trim();
 }
 
 export async function uploadFile(localPath: string, remotePath: string): Promise<void> {

packages/cli/src/hetzner/hetzner.ts

@@ -511,11 +511,13 @@ export async function runServer(cmd: string, timeoutSecs?: number, ip?: string):
 
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-
-  if (exitCode !== 0) {
-    throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+  try {
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server failed (exit ${exitCode}): ${cmd}`);
+    }
+  } finally {
+    clearTimeout(timer);
   }
 }
 
@@ -543,18 +545,20 @@ export async function runServerCapture(cmd: string, timeoutSecs?: number, ip?: s
 
   const timeout = (timeoutSecs || 300) * 1000;
   const timer = setTimeout(() => killWithTimeout(proc), timeout);
-  // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
-  const [stdout] = await Promise.all([
-    new Response(proc.stdout).text(),
-    new Response(proc.stderr).text(),
-  ]);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
-
-  if (exitCode !== 0) {
-    throw new Error(`run_server_capture failed (exit ${exitCode})`);
+  try {
+    // Drain both pipes before awaiting exit to prevent pipe buffer deadlock
+    const [stdout] = await Promise.all([
+      new Response(proc.stdout).text(),
+      new Response(proc.stderr).text(),
+    ]);
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`run_server_capture failed (exit ${exitCode})`);
+    }
+    return stdout.trim();
+  } finally {
+    clearTimeout(timer);
   }
-  return stdout.trim();
 }
 
 export async function uploadFile(localPath: string, remotePath: string, ip?: string): Promise<void> {

packages/cli/src/shared/ssh.ts

@@ -120,7 +120,7 @@ export function killWithTimeout(
   } catch {
     return;
   }
-  setTimeout(() => {
+  const sigkillTimer = setTimeout(() => {
     try {
       if (!proc.killed) {
         proc.kill(9);
@@ -129,6 +129,9 @@ export function killWithTimeout(
       /* already dead */
     }
   }, gracePeriodMs);
+  // Don't let this timer keep the event loop alive — the process may already
+  // be dead from SIGTERM, so there's no reason to block exit for 5 seconds.
+  sigkillTimer.unref();
 }
 
 // ─── TCP Pre-Check ───────────────────────────────────────────────────────────

packages/cli/src/sprite/sprite.ts

@@ -461,10 +461,13 @@ export async function runSprite(cmd: string, timeoutSecs?: number): Promise<void
     );
     const timeout = (timeoutSecs || 300) * 1000;
     const timer = setTimeout(() => killWithTimeout(proc), timeout);
-    const exitCode = await proc.exited;
-    clearTimeout(timer);
-    if (exitCode !== 0) {
-      throw new Error(`sprite exec failed (exit ${exitCode}): ${cmd.slice(0, 80)}`);
+    try {
+      const exitCode = await proc.exited;
+      if (exitCode !== 0) {
+        throw new Error(`sprite exec failed (exit ${exitCode}): ${cmd.slice(0, 80)}`);
+      }
+    } finally {
+      clearTimeout(timer);
     }
   });
 }
@@ -492,9 +495,15 @@ async function runSpriteSilent(cmd: string): Promise<void> {
       ],
     },
   );
-  const exitCode = await proc.exited;
-  if (exitCode !== 0) {
-    throw new Error(`sprite exec (silent) failed (exit ${exitCode})`);
+  // 60s timeout — silent commands should not hang indefinitely
+  const timer = setTimeout(() => killWithTimeout(proc), 60_000);
+  try {
+    const exitCode = await proc.exited;
+    if (exitCode !== 0) {
+      throw new Error(`sprite exec (silent) failed (exit ${exitCode})`);
+    }
+  } finally {
+    clearTimeout(timer);
   }
 }
 
@@ -629,8 +638,12 @@ export async function destroyServer(name?: string): Promise<void> {
   const stderrText = new Response(proc.stderr).text();
   // 60s timeout — sprite destroy should not hang indefinitely
   const timer = setTimeout(() => killWithTimeout(proc), 60_000);
-  const exitCode = await proc.exited;
-  clearTimeout(timer);
+  let exitCode: number;
+  try {
+    exitCode = await proc.exited;
+  } finally {
+    clearTimeout(timer);
+  }
   if (exitCode !== 0) {
     logError(`Failed to destroy sprite '${target}'`);
     logError(`Delete it manually: sprite destroy ${target}`);