Run Less, Prove More: Diff-Scoped Test Gating for Agent Patc

Order-based testing reveals that flaky tests in Agent Patcher are not truly random but suffer from shared state issues. While randomization hides this, bisecting test pollution identifies the root cause. This distinction is crucial: random failures require temporary freezes, whereas order-related failures need fixed repairs. A proposed solution for order-related failures involves diff-scoped test gating, which runs a subset of tests, justifies their inclusion, and escalates to a full suite if thresholds are exceeded. This approach aims to optimize speed while ensuring reliability.