openai/gpt-5.3-chat → A (3:1)

Side A meaningfully improves the testing infrastructure by replacing brittle, manually enumerated suites with automatic discovery, reducing maintenance overhead and preventing future test omissions. Side B is a small but important bug fix (adding a missing configuration variable), yet its scope and impact are limited compared to the broader structural improvement in Side A.