[BUG] Claude Desktop v1.1.5368 Windows: rootfs.vhdx detected as missing but download never initiates, app completely unusable

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?

What's Wrong?

Claude Desktop detects rootfs.vhdx missing on every VM startup attempt but never initiates the download. The app enters an infinite loop of loading the vmClient module, detecting the missing file, and doing nothing about it. The Downloading rootfs.vhdx... log line that should follow never appears.

This makes the entire desktop app unusable, not just Cowork. The app never gets past the initial loading state, so there is no error banner, no "share your debug logs" button, and no way to access Chat mode.

Environment

• Claude Desktop version: 1.1.5368.0 (MSIX, clean reinstall)
• OS: Windows 11, Build 26200
• CPU: Intel Core Ultra 7 268V
• RAM: 32GB (22GB free at time of failure)
• Hyper-V: Enabled

How It Started

Cowork was running normally. While switching between Cowork and Code interfaces, the Cowork session was terminated. Logs show a race condition during shutdown:

08:16:40 [info] [CoworkVMProcess] kill called with signal: SIGTERM
08:16:40 [info] [CoworkVMProcess] kill called with signal: SIGTERM
08:16:41 [info] [Process] Exited, code=0, signal=null, duration=373504ms, oom=false
08:16:41 [error] [CoworkVMProcess] kill failed with error: Error: RPC error -1: process already exited
08:16:45 [info] [CoworkVMProcess] kill called with signal: SIGKILL
08:17:02 [info] [startVM] VM already connected

The process exited cleanly (code=0) but the app sent a second SIGTERM to an already-dead process, got an RPC error, then escalated to SIGKILL. After this, [startVM] VM already connected indicates the VM layer believed the old connection was still live.

From this point forward, the app entered the stuck loop and never recovered.

What I Tried (All Failed)

Each step was performed after fully quitting Claude Desktop, in order:

1. Deleted VM bundles (%APPDATA%\Claude\vm_bundles and %APPDATA%\Claude\claude-code-vm), rebooted
2. Cleared caches (Electron Cache, Code Cache, claude-code directories), rebooted
3. Killed stale services (stopped all Claude/Cowork/vmwp processes, restarted CoworkVMService, vmcompute, vmms), deleted VM + cache directories, rebooted
4. Cleaned HNS network state (removed Cowork HNS network via Get-HnsNetwork | Remove-HnsNetwork), restarted Hyper-V services, deleted all VM + cache data, rebooted
5. Full MSIX uninstall + data wipe:

• Get-AppxPackage *Claude* | Remove-AppxPackage
• Deleted %APPDATA%\Claude entirely
• Deleted %LOCALAPPDATA%\Packages\Claude_pzs8sxrjxfjjc
• Rebooted, clean reinstall from claude.ai/download

1. Registry-level service deletion (CoworkVMService survived uninstall with Access Denied via sc.exe delete, required Remove-Item -Path "HKLM:\SYSTEM\CurrentControlSet\Services\CoworkVMService" -Recurse -Force), rebooted, reinstalled

None of these resolved the issue. After every attempt, the app returns to the same rootfs.vhdx missing loop.

Log Evidence

cowork_vm_node.log (post-reinstall, same pattern as pre-reinstall)

2026-03-08 09:13:16 [info] [VM] Loading vmClient (TypeScript) module...
2026-03-08 09:13:16 [info] [VM] Module loaded successfully
2026-03-08 09:13:19 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:13:19 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:18:40 [info] [VM] Loading vmClient (TypeScript) module...
2026-03-08 09:18:40 [info] [VM] Module loaded successfully
2026-03-08 09:18:42 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:18:42 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:18:42 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:18:42 [info] [warm] No warm file found for rootfs.vhdx at current version
2026-03-08 09:22:46 [info] [VM] Loading vmClient (TypeScript) module...
2026-03-08 09:22:46 [info] [VM] Module loaded successfully
2026-03-08 09:22:49 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:

[cowork_vm_node.log](https://github.com/user-attachments/files/25824454/cowork_vm_node.log)
[main.log](https://github.com/user-attachments/files/25824453/main.log)

22:49 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:22:49 [info] [Bundle:status] rootfs.vhdx missing
2026-03-08 09:22:49 [info] [warm] No warm file found for rootfs.vhdx at current version

Note: rootfs.vhdx not found, downloading... never appears. The download code path is never entered.

main.log (confirms clean environment post-reinstall)

current_version: '1.1.5368'
arch: 'x64'
os_version: '10.0.26200'
cpu_model: 'Intel(R) Core(TM) Ultra 7 268V'
total_memory: 34024411136
free_memory: 22375002112

What Should Happen?

Expected Behavior

After detecting rootfs.vhdx missing, the app should log rootfs.vhdx not found, downloading... and download the VM bundle, as documented in other users' working logs.

Impact

Claude Desktop is completely unusable. The app never reaches a functional state on any interface (Chat, Cowork, or Code). There is no error banner to click, no "share your debug logs" option, and no fallback to Chat-only mode.

Error Messages/Logs

Steps to Reproduce

1. Have a running Cowork session
2. Switch to the Code interface while Cowork is still active
3. Switch back to Cowork
4. Observe app hang and crash
5. Relaunch Claude Desktop
6. App never recovers, enters infinite rootfs.vhdx missing loop with no download attempt

After the initial crash, the broken state persists across reboots, full MSIX uninstall/reinstall, and manual deletion of all VM bundles, caches, and the CoworkVMService registry entry.

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

Claude Desktop: 1.1.5368

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

_No response_