Threshold Monitor Boundary Fuzzing

The threshold monitor starts a new window when timestamp >= window_start + time_window_secs. Its property test generates ordered success/failure sequences at end - 1, end, and end + 1.

For every sequence, the test compares check_thresholds with an independent current-window model. It asserts that the breaker opens exactly when the current window contains at least the configured number of failures. The test uses proptest, so a failing sequence is automatically shrunk to a small boundary-focused reproducer.

The generated inputs contain no token amounts and exercise no authorization paths. This keeps the security assertion focused on the monitor's window-membership boundary: failures before a rotation must not leak into the new window, while failures immediately before it must remain counted.