Environment credential issuance is unauthorized” when continuing an existing Codex cloud task

I’m using ChatGPT Business, Codex cloud environments and three AI assistants to work on the same private software project. Each assistant has its own account or environment context, and they coordinate their work through a shared GitHub repository. We want them to use consistent tools and configuration, but I have not verified that their cloud tasks run in the same live environment or under identical permissions.

The problem

When my main assistant, Willy 1, tries to continue an existing Account A cloud task, the request fails with:

“AppServerBackendRequestError: The caller does not have permission to execute the specified operation: environment credential issuance is unauthorized”

This happens when sending a new instruction to an existing task. The latest retry was a read-only request for an estimate, not a request to change credentials, install anything or write code. It returned the same error and did not start a new task turn.

What worked before

  • The affected Account A task previously ran successfully, wrote project files and later uploaded its review evidence. Its last successful turn is marked completed.
  • The assistant can still read that task’s recorded result. Reading it does not resolve the failure when trying to continue it.
  • Account C also reported the same credential-issuance error when trying to continue a task for a packaging step. I’m still collecting C’s exact task details and failure time.
  • Account B reported that an existing task continuation completed successfully at 06:13 UTC on 6 November, and another read-only continuation was admitted afterward. This suggests the issue is not affecting every existing cloud task, but it does not establish why A and C failed.
  • We have not tested whether creating a fresh task would produce the same error, so we cannot say whether the issue affects new tasks too.

What we’ve tried and observed

  • We retried the affected Account A continuation at my request, including a read-only instruction. The same error returned.
  • The previous successful turn remains marked completed; the failed continuation did not replace it.
  • The team can see saved cloud environment entries, but seeing an environment in the catalog has not established permission to launch or continue a task in it.
  • We have not changed credentials, workspace access, security settings or paid usage to work around the error. We have not established whether the separate environment/account contexts are the cause.

What I’d like help with

  1. What does “environment credential issuance” refer to in this situation? Is it task continuation authorization, workspace membership, environment ownership, GitHub connection, or something else?
  2. Which supported settings or diagnostics should I check for the affected account, workspace and saved environment?
  3. Can an existing Codex cloud task become inaccessible for continuation even though its history remains readable? If so, what are the supported recovery steps?
  4. Is there a supported way for separate members of a ChatGPT Business workspace to use the same saved Codex environment configuration, and does that mean they share a live environment or only the configuration?
  5. What information should I give Support to trace this, and are there known service-side causes for this error?