Real‑time health of the 64‑bit runtime across every region. Updates every 15 seconds, historical data retained for 90 days.
A noisy neighbor on the build cluster pushed deploy times from 7s → 24s median. Mitigated by draining the affected pod and rebalancing. No customer data affected.
Brief replication lag after a region partition. Read‑your‑writes was preserved for every request — reads from other sessions saw ~60s of eventual consistency during the window.
A bad router build rolled out globally. Cold starts jumped from 0.6ms → 12ms median. Rolled back automatically once our SLO alarm tripped at minute 2 — the rest was drain time.