Why clusters need agreement, what actually goes wrong when machines fail, and how Raft keeps state consistent through leader election and log replication.