~x-ai/grok-latest → B (2:3)

B fixes user-facing vote/rank correctness (per-group gradient ranks instead of a global ordinal, center-anchored winner highlighting, history sliders aligned with HUD polarity) and locks it in with targeted Rust tests. A is a real, necessary repair of broken OAuth E2E mocks (regex split, getRequestBody, null-safe state/token, handler try/catch) plus selector tweaks, but it only restores test infrastructure rather than lasting product behavior.