[BUG] VM workspace download loop — rootfs.vhdx download always times out (China, enterprise gateway)
Body:
### Environment
- OS: Windows 11 Home China 10.0.26200
- Claude Desktop: 1.9659.2 (also reproduced on 1.9255.0)
- Claude Code: 2.1.156
- Platform: win32 x64
- Network: enterprise gateway (custom-3p provider: gateway)
### Description
The VM workspace never starts —"Workspace still starting" hangs indefinitely. The root cause is that rootfs.vhdx
cannot be downloaded. The app enters an infinite loop: download →timeout →auto-reinstall →delete bundle →download
again →timeout...
### Steps to reproduce
- Install Claude Desktop on Windows 11 Home China
- Open the app
- Observe "Workspace still starting" never completes
### Expected behavior
VM workspace starts within ~1 minute after rootfs.vhdx download completes.
### Actual behavior
The download of rootfs.vhdx consistently fails with network errors. The bundle directory
(%LOCALAPPDATA%\Claude-3p\vm_bundles\claudevm.bundle) remains empty (0 bytes).
### Errors from cowork_vm_node.log
All download attempts fail within 20-30 seconds:
ERR_CONNECTION_TIMED_OUT(most frequent)ERR_CONNECTION_RESETERR_CONNECTION_CLOSEDERR_TIMED_OUTERR_NETWORK_CHANGED
Sample log entries:
rootfs.vhdx not found, downloading...
Downloading rootfs.vhdx...
[error] [VM:start] Startup failed: Error: Request error: net::ERR_CONNECTION_TIMED_OUT
[VM:start] Auto-reinstalling workspace after startup failure
[deleteVMBundle] Reinstall files deleted
This cycle repeated 30+ times over 2 days without success.
### Additional context
- Bundle version:
5680b11bcdab215cccf07e0c0bd1bd9213b0c25d vm_bundlesfolder: 0 bytes, only.auto_reinstall_attemptedpresentclaude-code-vmfolder: 230MB (incomplete)- No proxy/VPN was active during most attempts
- This CLI version of Claude Code works fine —only the desktop app's VM sandbox is affected
### Workaround attempted
- Clearing cache, rebooting, reinstalling —no effect
- The download server appears unreachable from this network environment
### Suggestion
Provide an offline/alternate download source for rootfs.vhdx, or add a longer timeout with retry with backoff. The
current 20-30s timeout is insufficient for multi-GB downloads on high-latency or throttled connections.
---
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗