[BUG] Cowork Windows: CoworkVMService crash (SCM event 7034) with no recovery action; claude.exe hangs against the dead service for 3+ hours (v1.11847.5.0)

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Environment

• Claude Desktop: 1.11847.5.0 (MSIX / WindowsApps install)
• OS: Windows 11 Pro 24H2 (Build 26100.4061)
• Service: CoworkVMService (cowork-svc.exe), LocalSystem, StartType=Automatic
• Hyper-V: full stack enabled (Microsoft-Hyper-V-All, VirtualMachinePlatform, HypervisorPlatform)
• VM is healthy when the service is up: vm_boot 287-541ms, total startup 8.8-13.9s, API REACHABLE

Timeline (2026-06-10, UTC+9)

• 08:05 Auto-update to 1.11847.5.0 (SCM 7040 "set to Disabled" -> 7045 "service installed";

this reinstall pattern occurred 9 times in the last 16 days due to update cadence)

• 11:22 SCM Event 7034: "The Claude service terminated unexpectedly (1st time)"
• 13:24 and 14:00 Application Hang (Event 1002): claude.exe stopped interacting with Windows
• 14:52 cowork_vm_node.log repeats:

[vm-client] Event resubscribe failed: Error: connect ENOENT \\.\pipe\cowork-vm-service
(service still down ~3.5 hours after the crash; no auto-recovery)

• 15:07 After a manual service restart, the VM starts normally (13.9s), again at 15:09 (8.8s)

Problems

1. CoworkVMService is installed without SCM failure/recovery actions, so a single crash

leaves Cowork unusable until manual intervention.

1. claude.exe does not detect the dead service: it hangs (two Application Hang events)

instead of showing an error or restarting the service.

1. Possible regression in 1.11847.5.0: the crash occurred ~3h after this update.

I will add a comment if it recurs on this version.

Expected behavior

• Service installed with failure actions (auto-restart on crash), or
• The app detects pipe ENOENT, restarts/repairs the service, and shows a clear error message.

Logs

• Attached: SCM events (7034/7040/7045), Application Hang 1002, cowork_vm_node.log excerpt.

cowork_vm_diag.txt

What Should Happen?

• The service should be installed with SCM failure/recovery actions, so a crash

auto-restarts it (e.g., restart after 5s / 10s / 30s), or

• Claude Desktop should detect the dead service (pipe ENOENT), attempt to restart or

repair it, and show a clear, actionable error message instead of hanging.

• Either way, a single service crash should not leave Cowork unusable for hours.

Error Messages/Logs

[System / Service Control Manager] 2026-06-10 11:22:38, Event 7034:
"Claude 서비스가 예기치 않게 종료되었습니다. 이것이 1번째입니다."
(EN: The Claude service terminated unexpectedly. This has occurred 1 time.)

[Application / Application Hang] 2026-06-10 13:24:49 and 14:00:11, Event 1002:
claude.exe version 1.11847.5.0 stopped interacting with Windows and was closed.

[cowork_vm_node.log] 2026-06-10 14:52:16-18 (repeating):
[vm-client] Event resubscribe failed: Error: connect ENOENT \\.\pipe\cowork-vm-service

Steps to Reproduce

The crash itself has occurred once so far (not yet deterministically reproducible).
The steps below document the incident and how to verify the unrecoverable state:

1. Windows 11 Pro, Claude Desktop 1.11847.5.0 (MSIX install), Cowork enabled,

full Hyper-V stack enabled.

1. Auto-update ran at 08:05 (service reinstalled: SCM event 7040 "set to Disabled"

-> 7045 "service installed").

1. ~3 hours later (11:22), CoworkVMService terminated unexpectedly (SCM event 7034).
2. No recovery occurred. The cause can be verified on any install (elevated PowerShell):

sc.exe qfailure CoworkVMService
-> no failure/recovery actions are configured for the service.

1. With the service down, using/relaunching Claude Desktop produced two Application

Hang events (Event 1002, 13:24 and 14:00), and cowork_vm_node.log kept repeating:
[vm-client] Event resubscribe failed: Error: connect ENOENT \\.\pipe\cowork-vm-service

1. The state persisted ~3.5 hours until manual recovery:

Restart-Service CoworkVMService
-> the VM then boots normally (8.8-13.9s total, API REACHABLE).

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

N/A (CLI not installed on this machine). This report concerns Claude Desktop Cowork on Windows: Desktop 1.11847.5.0. The Cowork VM bundles Claude Code SDK 2.1.170 per cowork_vm_node.log ("Installing SDK: ... version=2.1.170").

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

• This is a Claude Desktop / Cowork (GUI) issue, not a CLI issue. The Terminal/Shell

field value was selected only because the form requires one.

• Platform: paid Claude subscription used via Claude Desktop Cowork.
• Update cadence context: the service was uninstalled/reinstalled 9 times in the last

16 days (1.8555.2 -> 1.11847.5), each time via SCM 7040 "Disabled" -> 7045
"installed". Full SCM excerpts are in the attached cowork_vm_diag.txt.

• Locale: Korean Windows. Event messages in the attachment are in Korean; the key ones

are translated above.

• Happy to provide further logs or run diagnostics on request. I will add a comment if

the crash recurs on 1.11847.5.0.