Question

You store metrics/events in a Postgres table partitioned by day (range partitioning on `event_time`). Ingestion is steady but the current day's partition is the write hot spot — and over the past week, write latency on inserts has been creeping up and now spikes badly near peak, even though older partitions are idle. Index bloat on the active partition's indexes looks high, and `pg_stat_user_tables` shows almost all updates/HOT-update misses concentrated there. Autovacuum mostly leaves the hot partition alone during the day. Triage and mitigate.