ChatGPT/Codex macOS app repeatedly crashes – Dispatch Thread Soft Limit Reached (build 6570)
What version of the Codex App are you using (From “About Codex” dialog)?
26.810.41047 (6570)
What subscription do you have?
PLus
What platform is your computer?
Mac: Apple Silicon
What issue are you seeing?
ChatGPT/Codex desktop app crashes consistently approximately 2–5 minutes after launch.
Version: 26.810.41047 (6570)
macOS: 26.6.1 (25G76)
Mac: Apple Silicon
The app worked normally on Friday 14 August. The issue began on Monday 17 August.
Every launch now results in a crash after a few minutes.
Crash report contains:
Dispatch Thread Soft Limit Reached: 64 (too many dispatch threads blocked in synchronous operations)
There are repeated calls involving:
sky.node
RemoteHostedPIPContentService
AESendMessage
I disabled the Computer Use integrations for Chrome and Excel and completely restarted the application, but the crash still occurs.
The latest crash occurred approximately 2.5 minutes after launch.
What steps can reproduce the bug?
Simply by opening up the ChatGPT app and leaving it, after a few minutes it crashes
What is the expected behavior?
_No response_
Additional information
_No response_
2 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
I can reproduce what appears to be the same or a closely related issue on a newer ChatGPT Desktop build.
Environment:
ChatGPT Desktop 26.818.22352 (build 6872)
macOS 26.6.2
Apple Silicon
~/.codex is a physical directory, not a symlink
Observed behavior:
The ChatGPT/Codex desktop app becomes increasingly sluggish during intensive Codex usage.
Very large Codex conversations/task histories appear to increase the likelihood of the problem, but they do not seem to be the root cause.
The app has repeatedly shown an unexpected termination/restart state.
No corresponding ChatGPT/Codex .ips crash report was generated in the cases I checked.
Relevant Unified Log entries repeatedly show the Computer Use path:
RemoteHostedPIPContent CUAService bootstrap failed:
Error Domain=NSOSStatusErrorDomain Code=-1712
RemoteHostedPIPContent requesting managed CUAService reconnect
I also confirmed that the affected service process is:
SkyComputerUseService
The service starts normally, performs signature/trust checks, and the macOS TCC access request succeeds:
access request result was 1, timeout was 0
It then establishes an XPC connection to SkyComputerUseClient, followed by:
invalidated because the client process (...) either cancelled the connection or exited
Shortly afterwards the CUAService bootstrap failed / reconnect sequence occurs.
There are also repeated XPC interruptions involving ChatGPT and Codex:
XPC_ERROR_CONNECTION_INTERRUPTED
and occasionally:
com.openai.codex-spks
failed to do a bootstrap look-up:
xpc_error=[3: No such process]
I tested disabling the explicit SkyComputerUseClient "turn-ended" notify hook in config.toml. This does not appear sufficient to eliminate the underlying Computer Use/CUA behavior.
One additional data point: unlike some related reports, the top-level ~/.codex directory on this system is not a symbolic link, so the symlink trigger discussed in related issues does not apply here.
The symptoms and log path look very similar to the sky.node / RemoteHostedPIPContent / Computer Use failures described in this issue and related reports.
I can provide additional redacted Unified Log excerpts if useful.