Status: IN
Gossip-based failure detection is the single correctness bottleneck for the distributed cluster: replication, read repair, and hinted handoff all depend on its timeout-driven liveness set, which is bounded by cleanup to prevent unbounded membership growth.