[BUG] v1.1.2998.0 introduces severe instability - multiple critical failures on Windows`
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:
- OS: Windows 10
- Claude Desktop: v1.1.2998.0 (released Feb 13, 2026)
- Previous version: v1.1.2685.0
- Installation: Microsoft Store version
---
Summary:
Update v1.1.2998.0 introduced severe stability problems that make Cowork essentially unusable on Windows. Multiple critical failures occur within minutes of use, requiring constant restarts/reinstalls.
---
Issues Introduced by v1.1.2998.0:
Issue A: Folder Mounting Completely Broken
Symptom:
When attempting to select a working folder, the operation fails with:
This task didn't load properly
CLI output was not valid JSON. This may indicate an error during startup.
Output: sandbox-helper: failed to unmount host share
(virtiofs at /mnt/.virtiofs-root: invalid argument,
plan9 at /mnt/.virtiofs-root/shared: invalid argument)
Issue B: SDK Daemon Connection Failures
Symptom:
This task didn't load properly
sdk-daemon not connected
Frequency: Intermittent - occurs randomly after 1-5 successful requests
Impact:
- Cowork becomes unresponsive
- Requires Claude Desktop restart
- Loses work in progress
---
Issue C: Socket Communication Errors
Symptom:
This task didn't load properly
failed to write length: An operation was attempted on something that is not a socket.
Frequency: Intermittent - occurs after sdk-daemon error or folder mount attempts
Impact:
- Complete communication breakdown between Claude Desktop and VM
- Requires full reinstall to temporarily fix
---
Failure Pattern (Critical):
The update exhibits a consistent instability cycle:
- Initial state: Reinstall or restart Claude Desktop
- Brief functionality: Cowork works for 1-5 requests
- Failure: One of the above errors occurs
- Non-recoverable: Cowork remains broken until restart/reinstall
- Cycle repeats: Each restart provides only temporary relief
Each failure:
- Consumes user's message/usage limits
- Wastes time requiring restarts
- Makes Cowork unreliable for any real work
---
Regression Analysis:
Previous Version (v1.1.2685.0):
- ✅ Cowork worked reliably in temporary workspace mode
- ✅ No socket/daemon/mounting errors
- ✅ Stable operation (aside from VPN incompatibility - see separate report)
- ❌ Folder selection UI was greyed out (non-functional)
Current Version (v1.1.2998.0):
-
- ❌ Folder mounting completely broken (virtiofs/plan9 errors)
- ❌ SDK daemon randomly disconnects
- ❌ Socket communication failures
- ❌ Requires constant restarts/reinstalls
- ❌ Overall: Significantly worse than previous version
What Should Happen?
Cowork should be operational, stable and working folder selection should be available.
Error Messages/Logs
**Diagnostic Logs:**
VM starts successfully but then fails:
[VM:start] Startup complete, total time: 10028ms
[VM] Network status: CONNECTED
[VM] API reachability: REACHABLE
[After 1-5 requests, one of these occurs:]
sdk-daemon not connected
OR
failed to write length: An operation was attempted on something that is not a socket
OR
sandbox-helper: failed to unmount host share (virtiofs/plan9 errors)
Impact:
v1.1.2998.0 makes Cowork completely unusable for production work on Windows:
Cannot rely on folder access (broken)
Cannot rely on stability (random failures every few minutes)
Cannot complete multi-step tasks (failures interrupt work)
Wastes usage limits on failed attempts
Extremely frustrating user experience
Urgency:
This is a critical regression. The update should be:
Rolled back immediately, OR
Hotfixed urgently
Users are better off on the previous version (v1.1.2685.0) despite its folder selection UI bug.
Suggested Actions:
Immediate: Consider rolling back v1.1.2998.0 or marking it as unstable
Urgent: Investigate virtiofs/plan9 mounting regression
Urgent: Investigate SDK daemon connection stability
Urgent: Investigate socket communication failures
Testing: Implement more thorough Windows testing before releases
User Communication: Warn Windows users about instability in v1.1.2998.0
Steps to Reproduce
Steps to Reproduce:
- Open Cowork tab
- Click folder selection (no longer greyed out - UI fixed)
- Select any folder within
C:\Users\[username]\ - Observe virtiofs/plan9 mount failure
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
Version 1.1.2998 (1f1d4d)
Platform
Anthropic API
Operating System
Windows
Terminal/Shell
Windows Terminal
Additional Information
Diagnostic Logs:
VM starts successfully but then fails:
[VM:start] Startup complete, total time: 10028ms
[VM] Network status: CONNECTED
[VM] API reachability: REACHABLE
[After 1-5 requests, one of these occurs:]
sdk-daemon not connected
OR
failed to write length: An operation was attempted on something that is not a socket
OR
sandbox-helper: failed to unmount host share (virtiofs/plan9 errors)
Impact:
v1.1.2998.0 makes Cowork completely unusable for production work on Windows:
Cannot rely on folder access (broken)
Cannot rely on stability (random failures every few minutes)
Cannot complete multi-step tasks (failures interrupt work)
Wastes usage limits on failed attempts
Extremely frustrating user experience
Urgency:
This is a critical regression. The update should be:
Rolled back immediately, OR
Hotfixed urgently
Users are better off on the previous version (v1.1.2685.0) despite its folder selection UI bug.
Suggested Actions:
Immediate: Consider rolling back v1.1.2998.0 or marking it as unstable
Urgent: Investigate virtiofs/plan9 mounting regression
Urgent: Investigate SDK daemon connection stability
Urgent: Investigate socket communication failures
Testing: Implement more thorough Windows testing before releases
User Communication: Warn Windows users about instability in v1.1.2998.0
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗