Codex Work consumes quota on failed executions and simple prompts take 5–10 minutes

No source file, test, or code entry point is identified; begin with Feedback ID 019f869f-6905-7fc3-87e6-f5b2cfcdc84e and Support case 12296417, using the screenshots to investigate stalled executions and quota accounting. Done means simple prompts complete promptly and failed or stalled executions do not consume unexpected weekly Work quota.

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

ChatGPT Work (Web) – 5.6 Terra Minimal

What subscription do you have?

ChatGPT Plus + Work/Codex

What platform is your computer?

Windows 10 x64

What issue are you seeing?

Since yesterday evening, Codex Work has become almost unusable.

Very simple prompts (for example translating about 30 lines of text or asking "Are you working normally?") take between 5 and 10 minutes to complete.

During those executions the interface repeatedly displays messages such as:

• Optimizing the conversation
• Reconnecting 2/2
• Thinking...
• Execution during 6–10 minutes

The responses are either extremely delayed or sometimes never complete.

The main issue is that my weekly Codex quota continues to decrease during these failed or stalled executions.

Examples observed:

• Weekly quota dropped from 88% to 72% while the work never completed.
• Another single response that took more than 6 minutes consumed about 4% of my weekly quota.
• I recharged to approximately 99% yesterday evening and the quota rapidly disappeared despite almost no productive work being completed.

This is not related to large coding tasks.

It also happens on extremely simple prompts such as:

• "Translate this"
• "Are you still in zombie mode or are you working normally?"

I have multiple screenshots showing:

• execution times between 5 and 10 minutes,
• repeated "Optimizing the conversation",
• "Reconnecting 2/2",
• quota decreasing from 88% to 72%, then to 68%, then to 65%.

I already submitted ChatGPT feedback.

Feedback ID:
019f869f-6905-7fc3-87e6-f5b2cfcdc84e

OpenAI Support case:
12296417

Please investigate whether failed or stalled executions are incorrectly consuming Codex weekly quota and whether this behavior is caused by a backend issue.

I can provide all screenshots if needed.

What steps can reproduce the bug?

Feedback ID: 019f869f-6905-7fc3-87e6-f5b2cfcdc84e

What is the expected behavior?

Simple prompts should complete within a few seconds.

The interface should not remain stuck on "Optimizing the conversation" or "Reconnecting 2/2" for several minutes.

If an execution fails or stalls, it should not consume an unexpected amount of weekly Work quota.

Additional information

I have multiple screenshots showing:

• execution times between 5 and 10 minutes;
• repeated "Optimizing the conversation";
• repeated "Reconnecting 2/2";
• weekly Work quota decreasing from approximately 99% after the weekly reset, then 88%, 72%, 68% and finally 65%.

Examples observed:

• A translation of about 30 lines remained stuck for several minutes.
• A simple question ("Are you still in zombie mode or are you working normally?") took more than 6 minutes to answer.
• Another execution remained stuck for nearly 10 minutes.

I have already submitted in-product diagnostics through the Feedback tool.

PS : Additional evidence: This screenshot shows the ChatGPT Desktop application remaining in an "Optimizing the conversation" loop for more than 25 minutes, repeatedly showing Optimized the conversation and Ran command, while Work quota continues to decrease. No useful answer is produced during this period.

Feedback ID:
019f869f-6905-7fc3-87e6-f5b2cfcdc84e

OpenAI Support case:
#12296417

I also have screenshots documenting these executions and can provide them if needed.