Sustainable On-Call
An on-call arrangement whose alert volume, rotation size and compensation allow it to continue indefinitely without degrading the people in it.
The workable numbers, which are worth stating because they are frequently violated:
At most two pages per shift. Above that, responders cannot investigate properly, and alert quality collapses because everything is triaged rather than understood.
A rotation of at least six to eight people, so the frequency is tolerable and expertise stays spread. A three-person rotation burns out, and a one-person rotation is a resignation on a timer.
Compensation — payment or time off in lieu. Unpaid on-call is unpaid work, and it is where resentment originates.
Explicit follow-up time after a disrupted night. An engineer who was up at 03:00 should not be expected in a 09:00 planning meeting.
The structural requirement: the team that builds the service is on call for it. This is not a punishment but a feedback loop — the people who can fix the causes of pages are the people who feel them. Separating them removes the incentive to reduce alerts, and the alert volume then only grows.
Every page must be actionable and carry a runbook link. A page that requires no action is a defect in the alerting, and should be treated as one.