[Bug] Cannot switch to another OpenCode Go model after a regional availability error

[Oceanpie]

Affected area

• Coding agent and CLI
• TUI model selection

What happened?

I was using an OpenCode Go subscription with opencode-go/gpt-5.6-luna. A request returned a regional service-availability error.

I then selected opencode-go/deepseek-v4-flash. Prime Agent required authentication configuration before the selected model could make a request, even though the OpenCode Go subscription was already configured.

Steps to reproduce

1. Configure an OpenCode Go subscription in Prime Agent.
2. Select opencode-go/gpt-5.6-luna.
3. Send a request and receive a regional service-availability error.
4. Select opencode-go/deepseek-v4-flash through the model selector or /model.
5. Observe the authentication configuration requirement before any request reaches the selected model.

Expected behavior

Explicitly selecting a different model from the same configured provider should allow that model to make a request using the existing authentication. If the request succeeds, the provider should become available again. If it fails, the existing stale state should remain.

Prime Agent version

Prime Agent 0.7.2 (source commit 9bf49d897). Current main at 7ca44937f (0.7.3) has the same model-selection gate in source.