Wake up only for criticals

Three crons, three urgency levels. The disk-fills-in-7-days check fires urgency: low: it logs to history and nobody hears about it. The error-rate-doubled check fires urgency: normal: it banners the dashboard tab during work hours. The site-is-down check fires urgency: critical, expire_time: 0: it locks the screen on every device with the display tab open until someone dismisses it.

That's the whole on-call grammar. Three fields in the JSON body of a POST /notify. No escalation policy editor. No on-call rotation calendar to misconfigure. The cron decides the urgency. Your phone's OS treats the criticals like criticals because the browser surfaced them as web notifications, and the OS doesn't ask twice.

You used to pay a per-seat SaaS to get this rule right. Now the rule is a curl flag.