[OpenCode Go] Dashboard rolling usage does not match the actual remaining usage #36372

Description

The rolling usage displayed on the OpenCode Go dashboard does not match the actual usage enforced by the backend.

The dashboard indicates that there is 100% rolling usage, but requests are rejected with a GoUsageLimitError, indicating that the backend considers the 5-hour rolling usage limit to have already been reached.
Additionally, when sending a request to an OpenCode Go model directly through OpenCode, the request remains pending indefinitely without returning either a response or an error.

Example error returned through the GitHub Copilot Custom Provider:

Sorry, your request failed. Please try again.

Reason: Rate limit exceeded

{
  "type": "GoUsageLimitError",
  "message": "5-hour usage limit reached. Resets in 21min. To continue using this model now, enable usage from your available balance."
}

ChatRateLimited: Rate limit exceeded

This suggests that the dashboard and the backend are using different usage values, making it difficult to determine the actual remaining rolling usage.

Expected behavior

• The rolling usage shown on the dashboard should accurately reflect the usage used by the backend to enforce limits.
• If the backend rejects requests because the rolling limit has been reached, the dashboard should also indicate that the rolling limit has been exhausted.

Actual behavior

The dashboard indicates that usage is still available, while requests made through the GitHub Copilot Custom Provider are rejected with a GoUsageLimitError because the backend considers the rolling usage limit to have already been reached. During the same usage-limit state, requests sent to OpenCode Go models directly through OpenCode also fail to complete and remain pending without returning either a response or a visible error. This may be another manifestation of the same backend usage-state inconsistency, although OpenCode does not expose the underlying error.