[BUG] 2.1.206 Turning off Usage credits option returns completed but half finished workflows when workflows hit 5-hour usage limit

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

I spawned a workflow in my main session, and while the workflow was running, it hit the five-hour usage limit and proceeded to use my usage credits.

I went to my account and turned off the usage credits. Consequently, the sub-agents in the workflow that had not finished returned as failed.

However, the main session still proceeded to report back a completed workflow and notified me that I had hit the five-hour usage limit.

I believe without initially having activated the usage credits option, the workflow would have paused, and I could have typed a continue prompt after the five-hour usage limit resets to continue the workflow instead of having the report back as completed.

What Should Happen?

During a hot turnoff of usage credits, Claude Code should have suspended the workflow instead of failing the sub-agents, or having them run out of tokens and report back as failed.

This way, after the five-hour user limit resets, I can still manually inject a continue prompt to finish the workflow, instead of Claude Code closing the half-finished workflow and returning it as completed.

Error Messages/Logs

Steps to Reproduce

Set the account usage credit option to be on.

Start a workflow when the five-hour usage limit is close to being used up.

Make sure the usage credit option is still on, as the workflow uses up the remaining five-hour usage limit.

After the session is tapping into Usage Credit, go to your Claude account and turn off the Usage Credit toggle.

Momentarily, the session should start to fail. Or, more accurately, the sub-agents within the workflow should start to fail. Eventually, the main session will report back a completed workflow, but it will only be half-finished because not all sub-agents have returned completed sub-tasks.

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.206

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Other

Additional Information

My main session's model was set as Fable. Coding effort is set to UltraCode, and all the subsequently spawned sub-agents are also using Fable.

The terminal emulator I am using is Ghosty.