Cache failover clears hot data

Your read path is cache (Redis) in front of a database, ~95% hit rate normally. At 21:00 the managed Redis has a brief blip, it's only unreachable for 8 seconds, but it comes back EMPTY (the failover cleared it). For the next several minutes your database is melting: DB CPU 100%, connection pool exhausted, query latency 10x, app error rate 35%. Redis itself is now healthy and reachable; it's just cold. No deploy. The 8-second blip ended minutes ago. How do you triage and mitigate?