Test cluster resilience under broker failures — follower restart, leader restart with re-election, topic reconciliation, and broker failover with topic reassignment.
