Phase 8.5 — Alerting and observability

Until this phase, nothing on the new platform could tell anyone when something broke.

Backups reported success or failure into a mailbox that was never delivered anywhere, so a failed backup looked exactly like a successful one. The same was true of the backup server's own housekeeping.

This phase closes that: failures now reach a phone directly, and the measurements that predicted the outages which started this whole migration get monitored so they can never again be discovered weeks after the fact.

No downtime.