Codex usage quota exhausted unexpectedly after a small number of interactions

Summary

A Codex Desktop session appeared to exhaust the 5-hour usage quota unexpectedly after only a few user-visible questions in the conversation. The tasks were ordinary local MATLAB script editing/debugging and brief explanation requests, not long-running autonomous work from the user's perspective.

Expected behavior

A short sequence of simple/local coding questions should not consume the full 5-hour allowance, or the UI should clearly explain what background activity counted toward the quota.

Observed behavior

The user reported that the 5-hour quota was gone despite asking only about three relatively simple questions. From the session context, there were local file edits, MATLAB validation attempts, image/screenshot review, and several follow-up corrections, but the user-visible request count was small.

Sanitized context

• Environment: Codex Desktop on Windows.
• Work type: local MATLAB script editing for image-processing/realtime plotting.
• Local paths, dataset names, screenshots, and user-specific identifiers have been removed.
• The session included code edits, local shell/MATLAB validation, and figure snapshot generation.
• No credentials or proprietary data are included in this report.

Why this is confusing

The user judges usage by the number and complexity of their explicit questions. If hidden/background work, tool calls, retries, validations, or long context processing consume the same quota, the UI did not make that clear enough.

Request

Please investigate whether Codex Desktop quota accounting is expected in this scenario, and consider improving the UI to show:

1. How much quota each turn/tool run consumes.
2. Whether background validation/tool execution counts toward the 5-hour limit.
3. A clearer breakdown when the quota is nearly exhausted or exhausted.

Privacy note

This issue is intentionally sanitized. It omits exact local paths, source file names, dataset names, screenshots, and any project-specific details.