A parks visitation model breaches the municipal endpoint error-rate SLO. Which alert type should fire for on-call?

An operational page to on-call when endpoint error rates exceed the SLO is correct because a parks visitation model that breaches municipal endpoint reliability must wake responders immediately. Error-rate SLO alerts are operational reliability controls: they mobilize people who can roll back, scale, or fix serving issues. That alert path is distinct from—though coordinated with—an operational page to on-call when drift scores trigger retraining evaluation, which follows a slower model-quality cadence rather than signaling an endpoint error-rate breach. An operational page to on-call when weekly SLO trends are ready for review fails urgency and relevance for an SLO breach because paging for a routine weekly review creates alert fatigue and is not triggered by the actual error-rate breach. An operational page to on-call when drift scores trigger retraining evaluation fails because reliability incidents need on-call response for endpoint errors now; a drift-triggered retraining trigger is a model-quality signal, not evidence that the endpoint is erroring now. An operational page to on-call when data ingestion completes for nightly sync fails as a control because a routine successful sync is not an incident, and paging on-call for it creates noise while never detecting the error-rate breach itself. Exam caveat: distinguish operational SLO/error-rate pages from operational pages to on-call when drift scores trigger retraining evaluation. When the stem cites endpoint error-rate SLO breach, choose on-call operational alerting. Operational check: configure an alert rule on visitation-endpoint error rate against the municipal SLO, verify it pages the on-call roster (not for weekly SLO trends ready for review), document rollback steps in the runbook, and keep operational pages for drift-triggered retraining evaluation on a separate quality-governance track.