[Bug]: Play/resume ignores provider change after hitting Claude usage limit #17689

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/desktop

Steps to reproduce

1. Use Claude in an existing thread until the provider's usage limit is reached and the turn is paused.
2. In the same thread, change the composer provider/model to Codex/OpenAI (GPT-6.1-Sol in the reported case).
3. Press the Play/resume button to continue the interrupted task.

Expected behavior

The resumed request should use the newly selected provider/model and continue the task. A usage limit from the previous provider should not prevent continuation with another provider.

Actual behavior

Play sends the automatic message Continue where you left off. but the provider change is not applied: the thread reports the Claude usage limit again, even though GPT-6.1-Sol is selected in the composer. The thread remains limited instead of continuing with the selected provider.

Impact

Major degradation: cannot resume the interrupted task with another provider through Play after reaching the original provider's usage limit.

Version or commit

T3 Code (Nightly); exact build/commit unknown.

Environment

macOS desktop app. Original provider: Claude. Newly selected model: GPT-6.1-Sol (Codex/OpenAI).

Screenshot observations

The supplied screenshot shows:

• Claude usage limit reached. This turn is paused until the 5-hour limit resets in 1h 14m.
• The automatically sent Continue where you left off. message, followed by another Usage limit reached. response.
• GPT-6.1-Sol selected in the composer while the thread is still marked Limited.
• The composer banner says Usage limit reached Reset time unavailable; retry manually.

Workaround

No confirmed workaround.