openai/gpt-chat-latest → A (4:1)

Side A fixes multiple concrete failures in the OAuth test infrastructure that were breaking end-to-end authentication: it corrects query parsing (`str/split` with regex), reads POST bodies from `getRequestBody`, guards against null bearer tokens and missing state, fixes redirect response handling, wraps handlers to avoid crashes, and updates Playwright test selectors and alias handling. Side B improves the UI by coloring rank rows based on normalized score ranges instead of list position and adds focused tests, but it is primarily a presentation enhancement rather than a broad reliability fix.