[Desktop/macOS] Opening an active Browser task blocks the app for ~31 seconds

Open
#36,703 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
45/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Quiet
Tech stack
macos, rust

Research direction

Start by reproducing the task-navigation steps on macOS and reviewing the app-server logs alongside the main-thread jank snapshots. Investigate the interaction between Browser-session restoration and screen-recording reconciliation; done means opening the affected task leaves sidebar navigation and the rest of the Desktop UI responsive without repeated timeout stalls.

Written by the indexing model from the issue text.

Description

app browser bug performance
What version of the Codex App are you using?
  • Codex Desktop: 26.727.51351
  • Bundle build: 6119
  • Chromium: 150.0.7871.182
What subscription do you have?

Not included; this appears to be a local Desktop UI/runtime issue rather than an account-specific behavior.

What platform is your computer?
  • macOS 26.5.1 (build 25F80)
  • Apple Silicon / arm64
What issue are you seeing?

Opening one particular still-running task that has used the in-app Browser makes the entire Codex window unresponsive for about 31 seconds.

If I click another task while Codex is stalled, the click is not processed until the stall ends. The affected task is a single active turn with many Browser tool calls and a context-compaction marker.

This reproduced three times in the same session:

Opened affected task Navigation away processed Stall
08:08:26.605Z 08:08:58.463Z 31.858 s
08:10:59.627Z 08:11:30.892Z 31.265 s
08:12:02.691Z 08:12:33.981Z 31.290 s

Immediately after each stall, the Desktop log records:

[sampler-manager] main_thread_jank_snapshot mainThreadJankCountSinceLastSnapshot=1

During the stalled intervals, the app-server sidecar repeatedly records:

screen recording reconcile failed ... screen capture display enumeration timed out after 5 seconds
terminating child process safely (with TERM)
terminating child process forcefully (with KILL)
child process forcefully terminated

The timing and logs suggest screen-recording reconciliation or Browser-session restoration may be blocking a main/UI-thread path. That is a correlation, not a proven root cause.

What steps can reproduce the bug?
  1. Start a still-running Codex Desktop task that uses the in-app Browser.
  2. Navigate to another task.
  3. Click the Browser-using task in the sidebar.
  4. While the app is stalled, click a different task.
  5. Observe that the entire app remains unresponsive and the second navigation is not processed for about 31 seconds.
  6. Repeat; the same delay occurs again.
What is the expected behavior?

Opening any task should keep sidebar navigation and the rest of the Desktop UI responsive. Browser/screen-recording restoration should run asynchronously, time out once, or fail gracefully without blocking the app.

Additional information

Possibly related, but not exact duplicates:

  • #33169 — app freezes when accessing existing threads
  • #30624 — Browser plug-in / in-app browser freezes
  • #25744 — Computer Use helper accumulation and macOS TCC/WindowServer stalls

Task contents, repository/workspace information, local paths, URLs, credentials, and raw logs are intentionally omitted. A narrower redacted diagnostic extract can be prepared privately if maintainers specify what they need.

Dominant language
Rust
Stars
125k
Forks
19.5k
Avg merge
1m
Merged PRs (30d)
1k

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from openai/codex

All issues in openai/codex

Similar issues

More Rust issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.