The Night Watchman's Lantern: What the Rounds Teach Us About Coverage

Before digital pings, there was the slow, deliberate circuit of the night watchman. His job, in the husk of a factory or the corridors of an institution, was not to prevent disaster directly but to ensure it did not go unnoticed. He carried a clock-like device and a lantern, stopping at pre-determined stations to wind a key, proving his presence. This practice, known as the watchman’s tour, offers a starkly physical lesson in what we now call uptime monitoring and coverage: that the path you choose matters as much as the fact you’re checking at all.

The watchman’s route was a statement of priorities. It was a coverage map etched in footsteps and lamp light. A station by the coal chute, another at the main gate, a third near the furnace. Each key-turn was a health check, an assertion that a specific point in the system was, at that moment, stable. Miss a station, and a silent failure could begin to smolder in the dark. The integrity of the entire system—the security of the building—depended on the completeness and regularity of his route. It’s the same with our services: a health check on an API gateway tells you nothing about a silently failing database replica. The lantern only illuminates what it is pointed at.

The Gap Between Stations

But the watchman also understood latency, in his own way. The time between his rounds was the system’s acknowledged blind spot. A fire could start five minutes after he passed the boiler room, and it would have nearly an hour to grow before his lantern again fell upon it. This was an accepted risk, balanced against the resources of a single man. In our world, we often pretend this gap doesn’t exist. We set a one-minute check interval and believe we have “continuous” monitoring. But what of the failure that blooms and dies in that fifty-ninth second? The watchman’s log acknowledged the interstitial silence; our dashboards often gloss over it, creating a false sense of total awareness.

Most poignant, perhaps, was the protocol for when the watchman himself failed to report. If the station keys went unwound, an alarm would eventually sound—not for a fire or an intruder, but for the absence of the observer. This is the ultimate lesson in observability: your monitoring must be observable. A health check that goes silent must itself trigger a higher-order alert, or you are left in a state of profound uncertainty. Is the service dead, or is it just the watcher?

Today, our rounds are automated, endless, and blindingly fast. But the principle of the tour remains. We must design our routes with the same deliberate intent as the night watchman, knowing that every check is a strategic choice to illuminate one point and leave another in shadow. We must be humble about the latency between our passes. And we must always ensure that if our lantern goes out, someone, or something, is watching for that too. Reliability isn’t built on seeing everything at once—it’s built on the disciplined, repeated illumination of the right things, in the right order, all through the night.

Notes & further reading

A few pages I came back to while writing this: