›_ Claude Code Issues
Claude Code Issues / #51553

Claude Desktop 1.3561.0 Windows MSIX: Utility process crashes hourly, auto-restarts, identical hardware on 1.3109.0 is stable

Resolved 💬 3 comments Opened Apr 21, 2026 by netmingames Closed Apr 25, 2026

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?

Summary

After MSIX auto-update from 1.3109.0 to 1.3561.0 on 2026-04-21 08:16, the desktop app crashes and auto-restarts roughly every 45–60 minutes. Crashes are a mix of two signatures:

  • Silent — main.log ends abruptly with no exit marker, no beforeQuit, no Sentry event; next entry is Starting app { appVersion: '1.3561.0' } ~15–25 seconds later.
  • Sentry-captured — [error] Sentry caught: { type: 'Unknown', value: "'Utility' process exited with 'crashed'", stack: undefined } with an event ID, often accompanied by GPU process gone: reason: 'crashed' and one or more live sessions throwing query error: Claude Code process exited with code 4294967295.

A second machine with identical GPU/OpenGL signature but still on the old version 1.3109.0 shows zero crashes over 3 days of continuous use with heavier workload (more sessions, more processes, more RAM). The only difference we can identify is the app version.

Sentry event IDs from today (2026-04-21)

  • 7cdb49369412405d8c66b4f554e32c15 — 14:09:09 'Utility' process exited with 'crashed'
  • (additional events at 09:19:08, 10:29:09 — event IDs in logs, can supply on request)

Timeline today

| Time | Event | Log signature |
|---|---|---|
| 08:12 | App running normally on 1.3109.0 | — |
| 08:16 | MSIX auto-update to 1.3561.0 | MSIX detected: Claude_1.3561.0.0_x64 |
| 09:19:08 | Crash | Sentry + GPU process gone + 2× session exit 4294967295 |
| 09:26:53 | Restart (manual?) | preceded by beforeQuit: handler fired, going down |
| 10:29:09 | Crash | Sentry + session exit 4294967295 |
| 10:56:16 | Restart | silent (no marker in main.log before Starting app) |
| 11:59:26 | Restart | silent |
| 13:04:09 | Crash | session exit 4294967295 |
| 13:07:11 | Restart | |
| 14:09:09 | Crash | Sentry 'Utility' process exited with 'crashed' (event 7cdb4936…) |

Conversations typically survive the restart (sessions auto-resume), but in-flight tool calls are lost. Work is disruptive.

Environment

  • Claude Desktop — 1.3561.0.0 (MSIX, Windows Store auto-update channel, family Claude_pzs8sxrjxfjjc)
  • CCD binary — claude-code/2.1.111/claude.exe
  • Node — 24.14.0 (bundled)
  • OS — Windows 10 Home, Build 19045
  • CPU — Intel Core i7-6700 @ 3.40 GHz
  • RAM — 16 GB
  • GPU — NVIDIA GeForce GTX 980, current driver 32.0.15.6094 (2024-08-14)
  • gpu-info.json reports all GPUs active: false, glImplementationParts: "(gl=none,angle=none)", skiaBackendType: "None" — i.e. software rendering. This is not a new condition — the comparison machine below has the same signature.

Comparison machine (stable, same user account / organization)

| | Crashing (this report) | Stable |
|---|---|---|
| App version | 1.3561.0.0 | 1.3109.0 |
| GPU | GTX 980 | GTX 1070 |
| gpu-info.json signature | all active: false, gl=none,angle=none | identical — all active: false, gl=none,angle=none |
| glImplementationParts | (gl=none,angle=none) | (gl=none,angle=none) |
| Claude-Code sessions on disk | 813 / 43 MB | 929 / 93.9 MB |
| claude.exe processes | 12 | 32 |
| Claude RAM total | 2.7 GB | 7.7 GB |
| Resident AV | Avira active | Avira active |
| Restarts in 3 days | 6+ today alone | 2 (normal boot) |

The comparison machine has been running 1.3109.0 unchanged since 2026-04-18 with heavier workload and has not crashed. Whatever regression is in 1.3561.x relative to 1.3109.0 appears to be the sole delta.

Representative log snippet (Sentry-captured case, 14:09:09)

2026-04-21 14:08:28 [info] [CCD] Session local_e2216561-… workflow bxhkfqddr ended (active: 0)
2026-04-21 14:08:35 [info] Making remote MCP tool call: transitionJiraIssue
2026-04-21 14:08:40 [info] Remote tool call succeeded: transitionJiraIssue
2026-04-21 14:08:57 [info] [Stop hook] Query completed for session local_e2216561-…
2026-04-21 14:08:57 [info] [CCD CycleHealth] healthy cycle for local_e2216561-… (222s, hadFirstResponse=true)
2026-04-21 14:09:09 [error] Sentry caught: {
  eventId: '7cdb49369412405d8c66b4f554e32c15',
  type: 'Unknown',
  value: "'Utility' process exited with 'crashed'",
  stack: undefined
}
2026-04-21 14:09:12 [info] Starting app { appVersion: '1.3561.0', ... }

The query completed cleanly (healthy cycle 222s) — the utility-process crash occurs ~12 seconds after the session went idle, not under load.

Silent-restart case (11:59:26, no marker)

2026-04-21 11:59:03 [info] [EventLogging] Flushing 1 events
2026-04-21 11:59:26 [info] Starting app { appVersion: '1.3561.0', ... }

No beforeQuit, no Sentry event, no crashed entry — main.log just jumps from routine event-flush to Starting app. Windows Event Log (Application + System) has no corresponding Application Error or WER dump in the window either. This variant appears to be Electron main process being killed without its own crash-handler firing.

Ruled out

  • Avira antivirus — enabled on both machines
  • Session-folder size — stable machine has more
  • Process count / RAM — stable machine has ~3× more
  • GPU hardware acceleration — same (disabled) on both
  • Chrome-Extension pairing — disabled on the crashing machine
  • Scheduled-task-driven session spawn — disabled since 13:30 today, crash at 14:09 happened anyway

Ask

  1. Any known Utility-process regression in 1.3561.x compared to 1.3109.0?
  2. Is there a supported way to roll back to 1.3109.0 on MSIX without breaking Store updates permanently?
  3. Would additional logs help — Sentry queue exports, full main.log, crashpad reports (if you know where they go on MSIX)?

Happy to attach anything you need via the Sentry event IDs above.

---

Labels: bug, area:desktop, platform:windows, regression

What Should Happen?

Update

Error Messages/Logs

Steps to Reproduce

Summary

After MSIX auto-update from 1.3109.0 to 1.3561.0 on 2026-04-21 08:16, the desktop app crashes and auto-restarts roughly every 45–60 minutes. Crashes are a mix of two signatures:

  • Silent — main.log ends abruptly with no exit marker, no beforeQuit, no Sentry event; next entry is Starting app { appVersion: '1.3561.0' } ~15–25 seconds later.
  • Sentry-captured — [error] Sentry caught: { type: 'Unknown', value: "'Utility' process exited with 'crashed'", stack: undefined } with an event ID, often accompanied by GPU process gone: reason: 'crashed' and one or more live sessions throwing query error: Claude Code process exited with code 4294967295.

A second machine with identical GPU/OpenGL signature but still on the old version 1.3109.0 shows zero crashes over 3 days of continuous use with heavier workload (more sessions, more processes, more RAM). The only difference we can identify is the app version.

Sentry event IDs from today (2026-04-21)

  • 7cdb49369412405d8c66b4f554e32c15 — 14:09:09 'Utility' process exited with 'crashed'
  • (additional events at 09:19:08, 10:29:09 — event IDs in logs, can supply on request)

Timeline today

| Time | Event | Log signature |
|---|---|---|
| 08:12 | App running normally on 1.3109.0 | — |
| 08:16 | MSIX auto-update to 1.3561.0 | MSIX detected: Claude_1.3561.0.0_x64 |
| 09:19:08 | Crash | Sentry + GPU process gone + 2× session exit 4294967295 |
| 09:26:53 | Restart (manual?) | preceded by beforeQuit: handler fired, going down |
| 10:29:09 | Crash | Sentry + session exit 4294967295 |
| 10:56:16 | Restart | silent (no marker in main.log before Starting app) |
| 11:59:26 | Restart | silent |
| 13:04:09 | Crash | session exit 4294967295 |
| 13:07:11 | Restart | |
| 14:09:09 | Crash | Sentry 'Utility' process exited with 'crashed' (event 7cdb4936…) |

Conversations typically survive the restart (sessions auto-resume), but in-flight tool calls are lost. Work is disruptive.

Environment

  • Claude Desktop — 1.3561.0.0 (MSIX, Windows Store auto-update channel, family Claude_pzs8sxrjxfjjc)
  • CCD binary — claude-code/2.1.111/claude.exe
  • Node — 24.14.0 (bundled)
  • OS — Windows 10 Home, Build 19045
  • CPU — Intel Core i7-6700 @ 3.40 GHz
  • RAM — 16 GB
  • GPU — NVIDIA GeForce GTX 980, current driver 32.0.15.6094 (2024-08-14)
  • gpu-info.json reports all GPUs active: false, glImplementationParts: "(gl=none,angle=none)", skiaBackendType: "None" — i.e. software rendering. This is not a new condition — the comparison machine below has the same signature.

Comparison machine (stable, same user account / organization)

| | Crashing (this report) | Stable |
|---|---|---|
| App version | 1.3561.0.0 | 1.3109.0 |
| GPU | GTX 980 | GTX 1070 |
| gpu-info.json signature | all active: false, gl=none,angle=none | identical — all active: false, gl=none,angle=none |
| glImplementationParts | (gl=none,angle=none) | (gl=none,angle=none) |
| Claude-Code sessions on disk | 813 / 43 MB | 929 / 93.9 MB |
| claude.exe processes | 12 | 32 |
| Claude RAM total | 2.7 GB | 7.7 GB |
| Resident AV | Avira active | Avira active |
| Restarts in 3 days | 6+ today alone | 2 (normal boot) |

The comparison machine has been running 1.3109.0 unchanged since 2026-04-18 with heavier workload and has not crashed. Whatever regression is in 1.3561.x relative to 1.3109.0 appears to be the sole delta.

Representative log snippet (Sentry-captured case, 14:09:09)

2026-04-21 14:08:28 [info] [CCD] Session local_e2216561-… workflow bxhkfqddr ended (active: 0)
2026-04-21 14:08:35 [info] Making remote MCP tool call: transitionJiraIssue
2026-04-21 14:08:40 [info] Remote tool call succeeded: transitionJiraIssue
2026-04-21 14:08:57 [info] [Stop hook] Query completed for session local_e2216561-…
2026-04-21 14:08:57 [info] [CCD CycleHealth] healthy cycle for local_e2216561-… (222s, hadFirstResponse=true)
2026-04-21 14:09:09 [error] Sentry caught: {
  eventId: '7cdb49369412405d8c66b4f554e32c15',
  type: 'Unknown',
  value: "'Utility' process exited with 'crashed'",
  stack: undefined
}
2026-04-21 14:09:12 [info] Starting app { appVersion: '1.3561.0', ... }

The query completed cleanly (healthy cycle 222s) — the utility-process crash occurs ~12 seconds after the session went idle, not under load.

Silent-restart case (11:59:26, no marker)

2026-04-21 11:59:03 [info] [EventLogging] Flushing 1 events
2026-04-21 11:59:26 [info] Starting app { appVersion: '1.3561.0', ... }

No beforeQuit, no Sentry event, no crashed entry — main.log just jumps from routine event-flush to Starting app. Windows Event Log (Application + System) has no corresponding Application Error or WER dump in the window either. This variant appears to be Electron main process being killed without its own crash-handler firing.

Ruled out

  • Avira antivirus — enabled on both machines
  • Session-folder size — stable machine has more
  • Process count / RAM — stable machine has ~3× more
  • GPU hardware acceleration — same (disabled) on both
  • Chrome-Extension pairing — disabled on the crashing machine
  • Scheduled-task-driven session spawn — disabled since 13:30 today, crash at 14:09 happened anyway

Ask

  1. Any known Utility-process regression in 1.3561.x compared to 1.3109.0?
  2. Is there a supported way to roll back to 1.3109.0 on MSIX without breaking Store updates permanently?
  3. Would additional logs help — Sentry queue exports, full main.log, crashpad reports (if you know where they go on MSIX)?

Happy to attach anything you need via the Sentry event IDs above.

---

Labels: bug, area:desktop, platform:windows, regression

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

1.3561.0.0

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

_No response_

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗