[BUG] Claude Desktop 1.11187.1 (Windows MSIX) crashes ~18s after launch — main process rss reaches ~2.7GB then dies

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?

Claude Desktop crashes consistently ~15-20 seconds after every launch.
The window appears and renders, then the process dies. Crash loop repeats
on every relaunch. Started immediately after auto-update from 1.10628.2.0
to 1.11187.1.0 (June 5, 2026). Previous version worked fine.

Environment

• App version: 1.11187.1 (MSIX / Windows Store install, package Claude_pzs8sxrjxfjjc)
• OS: Windows 11
• Account: Enterprise/Team org, 12 enabled skills, 4 custom MCP servers

Log evidence (main.log)

• Startup completes normally (main_view_dom_ready ~1.3s)
• [SkillsPlugin] Starting skills sync / Found 12 enabled skills
• [event-loop-stall] main process blocked ... rss 1647MB → rss 2748MB within ~18s of launch
• Log cuts off, app restarts. Pattern repeats every launch.

What I tried (none helped)

• Cleared local-agent-mode-sessions (had 5,750 session folders / 670MB) —

confirmed loading with existingSessions=0, still crashes

• Full PC reboot
• Launching with --disable-background-networking (workaround from #48942)
• Disabling all custom MCP servers (renamed claude_desktop_config.json)

Expected

App should not exhaust memory at startup. Possibly related to #48942
(skills sync memory leak) and #61992, but the --disable-background-networking
workaround does not help in my case.

main.log attached.

What Should Happen?

The app should stay running after launch. Memory should not grow to ~2.7GB
and crash within ~18 seconds of startup.

Error Messages/Logs

No explicit error before death — log simply cuts off after memory spike, then app restarts.

Last lines before each crash (main.log):
[SkillsPlugin] Starting skills sync
[SkillsPlugin] Found 12 enabled skills
[event-loop-stall] main process blocked for 1065ms (total 1, cumulative 1065ms, rss 1647MB)
[event-loop-stall] main process blocked for 3898ms (total 2, cumulative 4963ms, rss 2748MB)
(log ends — app relaunches ~19s later)

Also recurring:
[error] Sentry caught: { type: 'Unknown', value: 'No message', stack: undefined }

Tried and did NOT help:
- Clearing local-agent-mode-sessions (5,750 folders / 670MB) — relaunched with existingSessions=0, still crashes
- Full PC reboot
- --disable-background-networking (workaround from #48942)
- Disabling all custom MCP servers (renamed claude_desktop_config.json)

Steps to Reproduce

1. Claude Desktop auto-updates from 1.10628.2.0 to 1.11187.1.0 (Windows MSIX / Store install)
2. Launch Claude Desktop
3. Window appears and renders normally (dom-ready ~1.3s)
4. Within ~15-20 seconds, main process memory grows to ~2.7GB (per event-loop-stall warnings in main.log)
5. App dies with no error dialog, then relaunches — crash loop repeats on every launch

Happens 100% of the time. Account has 12 enabled skills and 4 custom MCP servers,
but crash persists even with all MCP servers disabled.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

1.10628.2.0

Claude Code Version

N/A — Claude Desktop app bug, not CLI. Desktop app version: 1.11187.1 (MSIX/Store)

Platform

Other

Operating System

Windows

Terminal/Shell

Other

Additional Information

Desktop app (GUI) bug — crash loop started immediately after MSIX auto-update
on June 5, 2026. Possibly related to #48942 / #61992 (startup OOM), but the
--disable-background-networking workaround does not help in my case.
Crash persists with all custom MCP servers disabled and with local sessions cleared.
main.log available on request.

main.log