Applications

Where gossip is used in production:

Hashicorp Consul / Serf: gossip for service discovery and key-value storage. Each datacenter is a gossip pool. Cross-datacenter via WAN gossip.

Cassandra: nodes gossip about cluster membership, schema, token ownership. Schema changes propagate via gossip.

DynamoDB: internal cluster topology and metadata via gossip. Customer-facing reads/writes don't use gossip directly.

Apache Kafka: ZooKeeper for now (legacy); KRaft (Raft-based) replacing ZK. Some operational metadata still gossip-friendly.

Bitcoin / Ethereum: block + transaction propagation. Each peer forwards new blocks/txs to a small set of peers. Within seconds, the entire network has the block.

Discord (Cassandra-backed): gossip about cluster topology between hundreds of nodes.

Cloudflare: gossip for cluster membership across hundreds of POPs.

Operating systems (mostly desktop): mDNS / Bonjour for local network device discovery.

For SMALL-SCALE INFO: gossip is overkill. Use a leader (Raft, Paxos). For LARGE-SCALE eventually-consistent: gossip is hard to beat.