Codex Cloud - Post Setup Caching isn't working
Resolved 💬 10 comments Opened Nov 13, 2025 by samskiter Closed Mar 29, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
Cloud
What subscription do you have?
ChatGPT Business
Which model were you using?
N/A
What platform is your computer?
N/A
What issue are you seeing?
Every time we run a task (even updating an existing task by asking follow on) we see the FULL setup script (not the maintenance script). We have container caching turned on and are asking questions in well under 12 hours.
This makes a simple change to an existing session take >10 minutes
What steps can reproduce the bug?
Start a task or comment on an existing task.
What is the expected behavior?
Codex should reuse the cached container from the previous run and execute quickly.
Additional information
_No response_
10 Comments
see also: https://github.com/openai/codex/issues/4871
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Nope, that duplicate actually shows others having the same issue
Yes, I have this issue too and the linked issue #4871 doesn't address it at all. This is so bad that I've abandoned codex cloud - if the base image doesn't have what you need then you have no recourse.
Still very broken @pakrym-oai this must be costing you a fortune?
This was also reported in the forums almost 3 months ago: https://community.openai.com/t/how-to-properly-utilize-codex-vm-cache/1354885
So I'm guessing that this isn't a priority for OAI.
Codex cloud is currently unusable.
I believe this may be fixed. Anyone else having better results @OptiWhisperer ?
I see it only partially improved, @samskiter. I did this:
So this is still probably unusable for me. I've moved on to running multiple cli instances in an EC2 container.
This bug report hasn't received any upvotes or follow-up posts in four months. I think it is likely fixed, so I'm closing. If you're still seeing a similar problem, please let us know.