[BUG] Desktop app renderer OOM-crash loop when opening a large cloud session (600 MB → 5 GB in ~50 s)

Status Open
Reported on v2.1.286
Maintainer reply None cached
Activity 1 comment · opened Oct 5, 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

The Claude Desktop app crashes repeatedly (about every 100 s). The renderer process grows to ~5 GB and is killed with an out-of-memory error. Opening a large cloud session in the UI triggers it.

Environment

  • OS: Windows 11 Home 10.0.26200
  • RAM: 31.8 GB total, 18.7 GB free. Disk: 325 GB free.
  • Claude Desktop 2.19675.0 (MSIX)
  • Embedded Claude Code 2.1.286

Evidence

  • App diagnostic log (sentry/scope_v3.json): 'renderer' process exited with 'oom', exitCode -536870904 (0xE0000008), level fatal, 4 occurrences.
  • No crash dumps: the Crashpad folder only has settings.dat, and the logs folder is empty.
  • Main process (~255 MB) and Claude Code process (~238 MB) stay stable in every measurement. Only the renderer grows.
  • Local data is small: sessions 4 MB, IndexedDB 8 MB, cache 33 MB. The largest local transcript is 3.4 MB.
  • Restarting the app does not help, because it reloads the same session.
  • A local Claude Code /loop session was ruled out as the cause (archived, problem persisted).

Controlled test

Main renderer memory sampled every 10 s. I closed the app, reopened it, then opened a large cloud session.

| Time | Action | Renderer (MB) |
|---|---|---|
| 21:30:17 to 21:31:37 | App open, large session not opened | 518 to 716 (stable) |
| 21:31:47 | Opened the large cloud session | 1,088 |
| 21:32:07 | | 2,177 |
| 21:32:27 | | 2,939 |
| 21:32:37 | | 3,942 |
| 21:32:47 | | 5,068 |
| 21:33:07 | | 4,922 |
| 21:33:17 | Renderer died (OOM) and was recreated (new PID) | 800 |

Earlier measurements with a large session restored on startup showed the same loop: about 800 MB to 4.7 GB in ~80 s, then death, restart, repeat.

Expected behavior

Opening a large session should not make the renderer grow without bound.

Impact

The app is unusable while a large cloud session is open. Work running in the session on the server is not affected.

Workaround

Do not keep the large session open in the UI. Archive it when it finishes and continue in a new session with a summary.

What Should Happen?

Opening a large cloud session should not make the renderer process grow without bound. The app should stay stable (renderer around 500 to 700 MB, as it is before the session is opened) instead of reaching ~5 GB and being killed with an OOM error.

Error Messages/Logs

'renderer' process exited with 'oom'
reason: oom
exitCode: -536870904 (0xE0000008)
level: fatal
(4 occurrences in sentry/scope_v3.json)

Renderer memory after opening the large cloud session (sampled every 10 s):
21:31:47  1088 MB
21:32:07  2177 MB
21:32:27  2939 MB
21:32:37  3942 MB
21:32:47  5068 MB
21:33:07  4922 MB
21:33:17  renderer died (OOM), recreated with 800 MB (new PID)

Before opening the session (21:30:17 to 21:31:37): stable at 518 to 716 MB.
Main process ~255 MB and Claude Code process ~238 MB stayed stable throughout.

Steps to Reproduce

  1. Open the Claude Desktop app on Windows 11 (v2.19675.0, MSIX).
  2. Open a very large cloud session (long history) in the UI.
  3. Watch the memory of the renderer process (e.g. with Task Manager or Get-CimInstance Win32_Process filtered by --type=renderer).
  4. The renderer grows from ~600 MB to ~5 GB in about 50 seconds.
  5. The renderer is killed with 'oom' (exitCode -536870904), the UI restarts, and the loop repeats if the app reopens the same session.

Notes:

  • With the app open and the large session NOT opened, renderer memory stays stable at 518 to 716 MB.
  • Only the renderer grows. The main process (~255 MB) and the Claude Code process (~238 MB) stay stable.
  • Local data is small (sessions 4 MB, IndexedDB 8 MB, cache 33 MB), so it is not a local cache or history problem.
  • I cannot attach the session itself. It is a long-running cloud session that is still working.

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.286

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

_No response_

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗