[BUG] Cowork: "Failed to create bridge sockets after 5 attempts" on every session since 1.3036.0 (macOS 26.3.1)

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?

Every Cowork session fails immediately after sending a message with "Claude Code process exited with code 1" on Claude Desktop 1.3036.0 (cf521c) on macOS 26.3.1.

The VM itself boots healthy — cowork_vm_swift.log shows clean startup through "API reachability: REACHABLE" and "SDK installed successfully: version=2.1.111". The failure is in the spawned in-VM claude binary, which exits ~1.3–1.8 seconds after spawn with this stderr in cowork_vm_node.log:

    [vm-stderr] Error: Failed to create bridge sockets after 5 attempts

This appears to be the same underlying mechanism as #19503, #23746, and #11079 — socat bridge creation failing during sandbox setup.

Affected:

• Both new (isResume=false) and resumed (isResume=true) sessions
• Both claude-sonnet-4-6 and claude-opus-4-7
• Sessions with iCloud-mounted and with local working folders
• Multiple users on the same Mac (my account and my spouse's account)

NOT affected:

• Claude Code CLI (standalone install) — works fine, goes through a different code path
• claude.ai in the browser — works fine

What Should Happen?

Cowork should successfully spawn the in-VM claude process and respond to messages, as it did prior to the 1.3036.0 auto-update earlier today.

Error Messages/Logs

From ~/Library/Logs/Claude/cowork_vm_node.log — repeats on every session spawn, 100% consistently:

[Spawn:vm] id=62a7ae61 Spawn succeeded in 151ms
[vm-stderr 62a7ae61] Error: Failed to create bridge sockets after 5 attempts
[Process:62a7ae61] Exited, code=1, signal=null, duration=1435ms, oom=false

Across 10+ captured crashes, durations cluster tightly between 1246ms and 1761ms — consistent with a deterministic timeout (5 retries × ~250ms) rather than a variable race.

VM itself boots cleanly every time. cowork_vm_swift.log confirms all startup steps complete through "API reachability: REACHABLE" and "SDK installed successfully: version=2.1.111". The failure is in the bridge socket setup performed by the in-VM `claude` binary after spawn.

Host-side lifecycle record from main.log:
  unhealthy_reason: 'system_error'
  error_message: 'Claude Code process exited with code 1'
  error_category: 'process_crashed'
  had_first_response: false
  seconds_to_outcome: 2

Steps to Reproduce

1. Update Claude Desktop to 1.3036.0 (cf521c) dated 2026-04-16T16:17:33.000Z (auto-update)
2. Open any Cowork session (new or resumed)
3. Send any message
4. Observe "Something went wrong — Claude Code process exited with code 1"
5. Logs at ~/Library/Logs/Claude/cowork_vm_node.log show "Failed to create bridge sockets after 5 attempts"

100% reproducible since ~13:42 ET on 2026-04-16.

Claude Model

Not sure / Multiple models

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Earlier build of 1.x today (prior to the 2026-04-16T16:17:33Z auto-update). I do not have the exact previous build number recorded.

Claude Code Version

Claude Desktop 1.3036.0 (cf521c) 2026-04-16T16:17:33.000Z — in-VM SDK reports version 2.1.111

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Troubleshooting already attempted without success:

• Logout and re-login
• Clearing ~/.local/share/claude
• Full app restart
• Opening new sessions with Sonnet 4.6 and Opus 4.7
• Wiping the VM bundle: rm -rf ~/Library/Application\ Support/Claude/vm_bundles and rm -rf /tmp/claude-*, then allowing the app to download and rebuild a fresh bundle
• Switching working folder away from iCloud Drive to local disk

Possibly related existing issues:

• #19503 — same "Failed to create bridge sockets" error, identified socat bridge creation against MITM proxy ports as root cause
• #23746 — same error after auto-update to 1.1.2156, affecting all tasks including new ones
• #11079 — same error when TMPDIR is non-existent or non-writable

The consistent 1.3–1.8 second duration between spawn and exit, across 20+ attempts on different sessions/models, suggests a deterministic race or configuration issue in the bridge socket setup path as it exists in SDK 2.1.111 bundled with Desktop 1.3036.0.