Fixed cluster-wide instability caused by a bloated cluster database and repeated leader-election failures

What happened

The cluster's own configuration database had grown far past its healthy size. Once over that threshold it began failing to elect a leader, and without a leader the cluster cannot accept changes or reliably report the state of what it is running.

Impact

Intermittent instability across the platform on 2026-01-05 — services flapping, deployments not taking effect, and the management layer unable to give a straight answer about what was actually running.

Resolution

The database was compacted and defragmented, its size limits raised to match the real size of the platform, and automatic compaction scheduled so it cannot drift back to the same state unnoticed.