Skip to content

claude desktop #347

Description

@gustavodelarosa-eng

Title: Renderer process repeatedly crashes with OOM (reason: 'oom') — frequency escalating over ~3 days

Environment:

  • OS: Windows 11 Pro (10.0.26200)
  • Install: MSIX / Windows Store package
  • App versions observed with this issue: 1.46388.4, 1.49585.0, 1.52386.0, 1.52386.3 (persists across auto-updates)

Symptom:
The main webview renderer process crashes with an out-of-memory reason and restarts itself. From %LOCALAPPDATA%\Claude\logs\main.log:

Main webview render process gone: { reason: 'oom', exitCode: -536870904 }

Each crash is preceded by the renderer's own memory ballooning from ~2GB to 4-6GB within 1-2 minutes (visible in the periodic [process-memory] log lines), while total system memory is not under pressure (several GB free out of 32GB total).

Frequency (escalating):

  • 2026-09-09: 6 crashes over ~40 minutes (roughly every 5-8 min)
  • 2026-09-10: 30 crashes, including a burst of 26 crashes in 40 minutes (~one every 90 seconds, 12:09-12:49)
  • 2026-09-11: recurred; the app became unresponsive enough that I had to fully uninstall/reinstall to recover. During that recovery I saw a Windows "the file is being used by another process" message — possibly a child process from an earlier crash still holding a file handle, unconfirmed.
  • No occurrences at all earlier in the same log file (back to 2026-08-20) — looks like a regression starting ~2026-09-09.

Likely same class of bug as other Electron AI-coding-assistant tools:
Same exact exitCode: -536870904 / reason: oom signature is reported in other Electron apps, notably Cursor's forum reports of "repeated renderer OOM crashes during agent-loop workflows" and renderer memory "steadily growing during/after AI workflows" (https://forum.cursor.com/t/repeated-renderer-oom-crashes-during-agent-loop-workflows/156475), plus Signal Desktop (signalapp/Signal-Desktop#7189) and VS Code (microsoft/vscode#303526). This suggests the leak may be in how long agent/chat sessions accumulate memory in the renderer (transcript rendering, message history, etc.) rather than something Claude-specific — worth checking if long-running sessions correlate with the crashes on this end too.

Impact: During the worst bursts the app effectively reloads every ~90 seconds, making it unusable. Full recovery required uninstalling and reinstalling.

Happy to share more log detail if useful — trimmed identifiers/session data out of this report.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions