Staging breaks constantly. That is partly the point. If those breaks climb the same ladder as production login failures, on-call will mute everything, including the real ones.
Default recommendation
- Production → on-call, escalation, night pages for customer impact.
- Staging / preview / QA → Slack or email, business hours, no secondary escalation.
Same monitoring product is fine. Same escalation policy is not.
When staging might page
- Staging is customer-facing (rare, but some B2B demos live there).
- A shared database means staging risk is production risk. Fix that architecture; do not paper over it with pages.
- Release trains that block on staging health overnight with a willing owner
Even then: separate policy, clearer name on the alert ("STAGING"), and no level-3 manager climb.
Hygiene
Label monitors clearly. Route by policy, not by hoping humans read the URL. After false night pages from staging, demote immediately. Do not "remind the team to ignore those."
Production pages should feel rare and sharp (page vs notify). Staging noise belongs in the day job. Mixing them is how you burn the only resource that matters: attention at 3am.
