What is your cutoff for killing a bad Codex run?

One thing I’m still trying to reason about with Codex is when a run should be stopped rather than allowed to keep spending context, steps, and time.

Some failures are obvious: repeated test failures, the same edit being attempted multiple times, or the agent circling around the same error message.

But the harder cases are more subtle:

A run looks like it is making progress, but the diff keeps growing in the wrong direction.

It keeps adding abstractions instead of fixing the actual issue.

It repeatedly re-reads files without changing its plan.

It burns a lot of time on environment setup instead of the task itself.

For people using Codex regularly, do you set a manual cutoff after a certain number of failed attempts, time elapsed, or repeated behavior?

Or do you usually let the run finish and inspect the result afterward?