Repeated Renderer Crashes #47985

What version of the Codex App are you using (From “About Codex” dialog)?

OpenAI.Codex_26.917.9434.0_x64__2p2nqsd0c76g0 (Version 26.917.71314)

What subscription do you have?

ChatGPT Pro 200

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

Newer app: repeated renderer crashes

Recent app logs and the Windows crash event identify package OpenAI.Codex_26.917.9434.0_x64__2p2nqsd0c76g0. The UI executable is ChatGPT.exe.

The app's Crashpad reports folder contains 20 metadata records dated September 24, all with:

{"capture_kind":"crash","osarch":"x86_64","ptype":"renderer"}

Seventeen records have file modification times after 6 p.m. These are metadata timestamps, not independently decoded crash-dump timestamps. The four latest are:

Renderer crashes are strongly consistent with the reported interface resets. The retained sidecar records do not identify the failing instruction, underlying cause, or which renderer/webview was involved. No full dump was available in this reports directory.

Evidence location: %APPDATA%\Codex\web\Codex\Crashpad\reports\*_sidecar.json.

Related errors in the newer app's logs

Files in the September 24 and 25 UTC log directories contain:

• 101 entries saying Conversation state not found.
• 51 sa_server_request_failed entries with HTTP status 500. Sample entries for /conversations/batch say Request timeout.
• Six entries saying Thread catalog sync failed. Inspected examples include Authenticated desktop fetch failed (429).
• A conversation-list request returning HTTP 429 with Too many requests.

For example, the log at 2026-09-25 01:54:44 UTC (September 24, 8:54:44 p.m. local) records a conversation batch request timing out with status 500. At 02:05:20 UTC (9:05:20 p.m. local), chat-list synchronization fails with status 429. At 02:05:27 UTC, the app logs Conversation state not found again.

These entries show conversation loading and synchronization failures alongside the crashes. They do not establish whether the network errors caused the crashes or whether missing conversation state followed renderer recovery. The 429 entries concern conversation synchronization; they do not by themselves establish a model-message allowance being exhausted.

Evidence location: %LOCALAPPDATA%\Codex\Logs\2026\09\24 and %LOCALAPPDATA%\Codex\Logs\2026\09\25.

What steps can reproduce the bug?

Feedback ID: no-active-thread-01a0d669-29a7-72a0-8ff9-19c15792c474

What is the expected behavior?

No response

Additional information

No response