openai/gpt-chat-latest → A (3:1)

Side A restores broken end-to-end OAuth authentication tests by fixing concrete server mock failures: it corrects form/body reading (`getRequestBody`), query parsing, redirect handling, null safety, and wraps handlers to avoid crashes, while also improving test interactions with selectors. Side B mainly improves vote comparison highlighting and history presentation with associated logic and tests, but these are UI behavior refinements rather than fixing a foundational test infrastructure regression.