Cache stampede on aligned expiry

At 18:00 exactly, every minute on the minute, your homepage feed service spikes: Redis cache-hit ratio briefly craters, the Postgres origin sees a burst of identical heavy 'trending posts' queries, p99 latency triples for ~10 seconds, then recovers. This started after yesterday's deploy that introduced a 'trending' module cached with a 60-second TTL and a fixed expiry aligned to the wall clock. How do you triage and fix this?