Purchased credits consumed at implausible rate inconsistent with Work usage and analytics

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

26.908.70816

What subscription do you have?

ChatGPT Pro

What platform is your computer?

macOS, Apple Silicon (arm64)

What issue are you seeing?

My purchased credits were consumed extremely rapidly while using ChatGPT Work, at a rate that appears fundamentally inconsistent with the work actually performed. This was not a particularly large or complex task, yet a substantial amount of purchased credit disappeared within a very short period.
The usage analytics also appear inconsistent with the amount of credit deducted, which is why I believe this may be a usage metering or billing bug rather than legitimate consumption.
I had been using Work increasingly frequently and it was becoming part of my daily workflow. I am now effectively unable to continue using it because I am not prepared to purchase further credits while I cannot account for how the existing credits were consumed.

What steps can reproduce the bug?

Feedback ID: no-active-thread-01a0aa0f-6468-7f71-bf78-dfacb124ec66

What is the expected behavior?

Credits should be deducted in accordance with the actual billable usage generated by the Work task and the published credit rates. The amount deducted should also be reconcilable with the usage information available to the user.
A relatively straightforward task should not unexpectedly consume a very large quantity of purchased credits without the usage data providing an explanation for that consumption.

Additional information

I was using ChatGPT Work normally on a relatively straightforward task. My included allowance was exhausted and purchased credits then began being consumed at an extremely rapid rate. Checking the usage analytics afterwards showed usage that does not appear capable of accounting for the amount of credit deducted.
I cannot safely attempt to reproduce this again because doing so would require purchasing further credits and risking losing them in the same way.
I have already raised this with OpenAI Support. The responses so far have been confused and have not explained what caused the usage, what is being investigated or, most importantly, what is being done to restore the credits or refund the money if they were incorrectly charged. The responses also appear to be AI generated, so I am not clear that the issue has yet reached a human reviewer.
This is particularly concerning because OpenAI's own usage analytics appear to support my contention that the amount deducted cannot be reconciled with the work performed.
There appear to be at least two existing reports describing very similar behaviour: #36622 and #45688. Issue #36622 reports approximately $90 of credits being consumed in around six hours despite comparable previous workloads consuming significantly less, with the usage dashboard unable to explain the consumption. Issue #45688 describes similarly extreme depletion of included usage followed by purchased credits.
My experience appears to be another independent instance of the same or a closely related problem. Until the cause has been identified and the disputed credits restored or refunded, I am unwilling to purchase further credits.

Start by comparing this report with issues #36622 and #45688 and the supplied Feedback ID no-active-thread-01a0aa0f-6468-7f71-bf78-dfacb124ec66. Reconcile the purchased-credit deductions with the reported usage analytics without requiring another paid reproduction; done means identifying the metering discrepancy and determining whether credits should be restored or refunded.