Developer Finds Wall-Clock Bug in Python Poller After 48-Hour Production Incident

A software developer spent 48 hours debugging a Python polling service that passed health checks locally but failed in production after deployment. The root cause was a deadline anchored to time.time(), a wall clock susceptible to NTP corrections and hypervisor pauses, rather than a stable monotonic clock. The developer refactored the wait_until helper to accept an injectable Clock abstraction, separating wall time from monotonic time. A FakeClock test fixture was built to simulate clock jumps deterministically inside pytest, eliminating the need to wait for real NTP events to reproduce the bug. The incident highlights why timeout logic should rely on monotonic clocks and be tested with controllable fake clocks rather than real system time.