A delivery service assigns the best available driver (a "dasher") to an order, but users report incorrect assignments. The root causes in the old implementation were a time-unit bug (timestamps are in milliseconds, while the freshness threshold is given in seconds) and inconsistent tie-breaking when GPS jitter makes multiple drivers appear equally close. Implement the corrected driver selector. Each dasher is represented as a 4-tuple: (driver_id, distance_meters, last_update_ms, available) Selection rules: 1. Ignore any dasher that is unavailable, has missing distance/last update (None), or has a negative distance. 2. A location update is stale if its age is greater than stale_after_seconds * 1000. If last_update_ms is in the future because of clock skew, treat its age as 0. 3. Among all valid dashers, find the minimum distance d_min. 4. Because of GPS jitter, any dasher with distance <= d_min + jitter_tolerance_m is considered equally close. 5. From those equally close dashers, choose the one with the most recent last_update_ms. 6. If there is still a tie, choose the dasher with the smallest driver_id. 7. If no valid dasher exists, return -1. In a real debugging interview, you would also discuss minimal failing tests, observability, retries, timeouts, and circuit breakers. For this coding exercise, implement the corrected selector function.