[BUG] Claude Desktop 1.17377.2 (Windows MSIX): event-loop-stall explosion after auto-update — session switching freezes UI for tens of seconds
Environment
- App: Claude Desktop (Windows Store / MSIX)
Claude_1.17377.2.0_x64__pzs8sxrjxfjjc - Bundled CLI (CCD): 2.1.197
- OS: Windows 11 Pro 10.0.26200
- Usage pattern: multiple parallel local sessions across several projects (13 warm
claudeCLI processes, ~4 GB RAM total)
Summary
After the app auto-updated to 1.17377.2 on 2026-07-02 at 15:16 (JST), the main process started logging massive numbers of [event-loop-stall] warnings. Switching between sessions now freezes the entire UI for tens of seconds. The previous version on the same machine, same data, was fine.
Evidence from %APPDATA%\Claude\logs\main.log
Daily aggregation of [event-loop-stall] main process blocked for Nms:
| Date | Stall count | Max single stall | Cumulative blocked time |
|------|------------|------------------|------------------------|
| 2026-07-01 (before update) | 6 | 885 ms | ~4 s |
| 2026-07-02 (updated 15:16) | 1,652 | 8,118 ms | ~48 min |
| 2026-07-03 (first ~40 min) | 24 | 6,022 ms | ~65 s |
The update event in the log:
2026-07-02 15:16:19 [info] [CCD] Initialized with version 2.1.197
2026-07-02 15:16:19 [info] [updater] MSIX detected: windowsStore=true, appPathMatch=true, source=windowsStore, exe=C:\Program Files\WindowsApps\Claude_1.17377.2.0_x64__pzs8sxrjxfjjc\app\Claude.exe
2026-07-02 15:16:27 [warn] [event-loop-stall] main process blocked for 1350ms (total 1, cumulative 1350ms, rss 333MB)
Stalls cluster directly after session-switch operations. Example:
2026-07-02 19:38:17 [info] [CCD] LocalSessions.setFocusedSession: sessionId=local_583ed280-...
2026-07-02 19:38:17 [info] LocalSessions.startShellPty: sessionId=local_583ed280-..., cols=80, rows=24
2026-07-02 19:38:21 [warn] [event-loop-stall] main process blocked for 5205ms (total 140, cumulative 172463ms, rss 556MB)
Consecutive multi-second stalls chain into the observed 10–30 s UI freezes on session switch.
Failed to set title bar overlay, this is probably expected also appears repeatedly, as reported in #68364.
Mitigations tried
- Deleted ~11,000 stale transcript files under
~/.claude/projects(2.4 GB → reduced) and setcleanupPeriodDays: 14. This reduced enumeration weight but stalls continue on 07-03 (24 stalls in the first 40 minutes of uptime), so data volume is an amplifier, not the root cause. - MSIX install: cannot pass Electron flags (
--disable-gpuetc.), no compatibility tab, downgrade not practical via Store.
Possibly related
- #68364 (MSIX, event-loop stalls during background VM bundle download + title bar overlay failures —
vm_bundlesis 9 GB on this machine) - #65550 (MSIX, event-loop-stall + memory growth after auto-update; only downgrade helped)
- #26302 (earlier precedent: severe UI lag regression right after desktop auto-update)
- #72725 (same 1.17377.x line, Windows-only spawn bug)
Expected behavior
Session switching should not block the main process for multiple seconds; an auto-update should not introduce a ~275x increase in event-loop stalls on the same machine and data.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗