Model Switch modal delays on long chats; opens only after fetch

Note: This is primarily a UX responsiveness issue rather than a functional defect; if your workflow classifies UX regressions as bugs, please track it accordingly.

In long chats, opening the model switch modal is delayed until a network fetch completes. In short/new chats, the modal appears instantly. This report is about UX impact and observed behavior, not a root-cause claim.

Repro steps

• Open a chat with many messages.

• Click to open the model switch modal.

• A network request (/api/trpc/getRateLimitData) fires and the modal renders only after the response.

Expected

Open the modal instantly with cached/minimal data or optimistically. Load the rest asynchronously so switching feels immediate, ideally even, continue loading after the user selects a model and begins typing, rather than blocking the modal from opening.

Actual
Modal rendering is blocked by the fetch. In large chats, the fetch is slower, creating a very noticeable 300ms > delay before the modal appears after the click.