Static membership holds partitions hostage when consumer fails repeatedly

When group.instance.id is configured, a consumer that exceeds max.poll.interval.ms or crashes preserves its static ID mapping. The Group Coordinator holds assigned partitions in reserve for session.timeout.ms, refusing to reassign them to healthy consumers. If the failure is deterministic (poison pill message), the same consumer reclaims the same partitions and fails again indefinitely.

Identify the group.instance.id associated with lagging partition. Change the group.instance.id (append timestamp or UUID) and redeploy, or use Kafka Admin API to explicitly delete the group member metadata. This forces the Coordinator to release the partition to other consumers. For deterministic failures, implement circuit breakers to skip poison pill messages.