Package main demonstrates heartbeat timeout and progress reporting.
Scenario: A long-running data processing job reports progress (percentage)
while running. The scheduler detects stale runs using a per-job heartbeat
timeout (10s) and auto-recovers if the node crashes.