Improve GPT-5.6 Ultra session-limit UX for long-running Goal/Codex workflows

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

26.707.31428

What subscription do you have?

Pro

What platform is your computer?

macOS

What issue are you seeing?

Summary

During a long autonomous coding session with GPT-5.6 Ultra, I reached the maximum continuous session duration. The product correctly prevented the session from continuing, but the recovery experience could be clearer for users running multi-hour engineering tasks.

Observed behavior

GPT-5.6 Ultra session reached its maximum duration.
The UI explained that the model was temporarily unavailable.
The client continued attempting to reconnect to GPT-5.6 Ultra.
It did not silently downgrade to GPT-5.5, which I believe is the correct behavior for engineering consistency.

Suggested improvements

Clearly distinguish:
Session duration limit
Weekly usage quota
Show an explicit countdown, for example:

GPT-5.6 Ultra session limit reached.

Available again in 39 minutes.

Explain why the session stopped:

Your session reached the maximum continuous runtime.

instead of implying a generic availability issue.

Offer explicit user choices:
Wait and resume with GPT-5.6 Ultra (recommended)
Switch this project to GPT-5.5 (results may differ)

Do not silently change models.

For Goal/Codex workflows, reassure the user that checkpointed work is safe and can be resumed.

Why this matters

Long-running engineering sessions are fundamentally different from ordinary chat sessions. Developers often spend several hours building a single project. Clear messaging about why execution paused, when it can resume, and what options are available reduces confusion and helps preserve confidence in long-running autonomous workflows.

What steps can reproduce the bug?

Hit the 5 hour limit and it tries to keep reconnecting instead of outright failing or dropping down to a lower model.

What is the expected behavior?

Stop trying to reconnect

Additional information

_No response_