Codex macOS app crashing on long tasks

Open 💬 1 comment Opened Apr 23, 2026 by sudo-relax

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

Version 26.417.41555 (1858)

What subscription do you have?

Pro

What platform is your computer?

Darwin 25.3.0 arm64 arm

What issue are you seeing?

The Codex desktop app has become increasingly unstable after recent updates. currently on Version 26.417.41555 (1858)

During long-running tasks (around 40–50 minutes), the app frequently crashes, especially when using OMX plugins. This has been a consistent issue over the past week.

The main problem is that when the app crashes, the task cannot resume from where it left off. This makes long workflows extremely frustrating and time-consuming, as there’s no clear way to recover progress or determine where the process failed.

Would appreciate a fix or at least a way to resume interrupted tasks.

What steps can reproduce the bug?

Steps to reproduce

  1. Open the Codex macOS desktop app.
  2. Start a long-running task that involves multiple tool calls and OMX plugin usage.
  3. Let the task continue for 40–50+ minutes.
  4. Keep the session active while Codex continues processing a large context and working through the task.
  5. After prolonged usage, the app eventually crashes unexpectedly.
  6. Reopen the app and try to continue the same workflow.
  7. Notice that the task does not reliably resume from the exact stopping point, making it hard to recover progress.

Bug explanation
The Codex macOS desktop app becomes unstable during long-running, high-context tasks, especially when OMX plugins are involved. After extended use, the app crashes unexpectedly. The bigger issue is not just the crash itself, but that recovery is poor: the interrupted task does not cleanly resume from the exact point where it stopped.

This creates major problems for longer workflows because:

  • significant progress can be lost
  • it is difficult to determine the last completed step
  • resuming requires manual inference and repetition
  • long sessions become unreliable for production use

What is the expected behavior?

_No response_

Additional information

_No response_

View original on GitHub ↗

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