Every distributed system you build is already taking a side in the CAP trade-off. The question is whether you made that choice deliberately or discover it during an incident.

CAP states that a distributed system can guarantee at most two of three properties: Consistency , Availability , and Partition Tolerance . The critical insight most teams miss — P is not optional. Networks fail. Pods crash. AZs go dark. You are choosing between CP and AP . Full stop.

CP vs AP: What You're Actually Trading

CP systems (etcd, ZooKeeper, CockroachDB) refuse to serve requests during a partition rather than return stale data. Leader-based consensus ensures correctness. Choose CP for financial ledgers, inventory reservation, distributed locks — any domain where stale reads are more dangerous than errors.

AP systems (Cassandra, DynamoDB, DNS) continue serving requests during a partition, accepting diverging state. Reconciliation happens later.…