Overprovisioned Resilience During Low-Risk Period

The previously-declared high-risk period has ended. Nothing is currently broken — every availability and latency metric is healthy — but the platform is still running the expensive resilience posture sized specifically for that now-concluded window: full active-active multi-AZ capacity, peak-sized failover reserve, high-frequency backups, and peak-event compute, standby, and cache headroom. Five business invariants govern every decision here: (1) the minimum redundancy floor must be preserved; (2) single-zone failure tolerance may not be removed if policy requires it; (3) peak-event-only capacity may be reduced after the event; (4) risk posture may be lowered only to the stated operating baseline; (5) cost savings are valuable, but never at the expense of mandatory resilience. The central judgment this scenario forces: a configuration can look perfectly healthy right now while genuinely sitting below the mandatory resilience floor, because the exposure that floor protects against is a real but infrequent event — a periodic, scheduled single-AZ failure drill — not something today's dashboard will ever show on its own. Decide which capacity is genuinely peak-event-only waste versus which capacity is the mandatory floor itself, reduce the former to the stated baseline, and never let aggressive cost-cutting remove the latter merely because nothing has failed yet.

Difficulty

Staff

Expected Time

~15 min

Category

Infrastructure & Capacity