The Watchtower's Unrung Bell: On the Fetish of the Silent Pager

There is a sacred object in the culture of reliability engineering, a golden calf we bow to in our daily stand-ups and quarterly reviews. It is the silent pager. The unwoken on-call engineer. The unblemished SLO. We speak of it in reverent tones: “My pager hasn’t gone off in weeks.” We chase it as the ultimate sign of a well-tended system. But I want to posit a uncomfortable, heretical idea: the fetishization of silence is a dangerous form of complacency. It mistakes the absence of alarms for the presence of health.

Our obsession stems from a reasonable place. The scream of a pager at 3 AM is a visceral, punishing experience. It represents a breach, a failure of our careful constructs. To seek relief from that pain is human. So we build thicker walls, add more checks, tune our thresholds ever tighter, all in pursuit of that blissful, uninterrupted sleep. We come to believe that a silent monitoring system is a successful one. But a watchtower whose bell never rings is not necessarily a sign of enduring peace. It could be that the bell has rusted shut, that the watchman has fallen asleep, or that the approaching threat is of a kind the bell was never cast to sound.

This is the insidious risk. In our drive to eliminate ‘noise’—those pesky, non-actionable alerts—we risk designing systems that can only see the catastrophes we’ve already imagined. We create a map of known failure modes and celebrate when nothing on that map lights up. Meanwhile, the service evolves. Dependencies shift. Usage patterns change in subtle ways. Novel, slow-growing pathologies emerge that don’t trip our binary thresholds but erode performance, user trust, and architectural integrity. The pager stays silent, and we take that as confirmation that all is well, while the foundation quietly settles into a new, less stable shape.

The Hum of Normalcy

A truly observable system shouldn’t be judged by its silence, but by the quality of its hum. Is it telling us a rich, continuous story about its state? Are we listening to the metrics of normal operation—the rhythm of request latencies, the melody of cache hit ratios, the cadence of garbage collection—with the same attentiveness we reserve for the klaxon of a 500 error? The silence we should seek isn’t the absence of all alerts; it’s the absence of surprise. It’s the confidence that comes from understanding the system’s baseline so intimately that any deviation, long before it becomes page-worthy, is noticed, understood, and contextualized.

Chasing the unrung bell, then, becomes a form of operational vanity. It encourages us to build systems optimized for a metric—mean time between failures—rather than for genuine resilience and understandability. It pushes us to suppress signals in the name of peace, rather than to expand our capacity to interpret them. The goal should not be a pager that never vibrates, but a team that is never bewildered by why it did. Sometimes, the most reliable service isn’t the one that never pages you; it’s the one that pages you for the right, subtle reason at 2 PM on a Tuesday, giving you the gift of foresight. That is a bell worth listening for.

Notes & further reading

A few pages I came back to while writing this: