Things that separate professional on-call from chaotic on-call:
- Clear primary and secondary — one person on point, one as backup
- Handover at end of shift — outgoing on-call briefs incoming on any ongoing issues, recent changes, anything weird
- Runbooks for common alerts — every page should have a runbook entry; if it doesn't, that's an action item
- Alert hygiene — paged alerts must be actionable; non-actionable alerts become tickets, not pages
- Escalation paths — when to wake the senior, when to call the manager, when to declare an incident
- No heroic on-call — if the system requires constant manual intervention, that's a reliability problem, not an on-call problem