Eventual Consistency
Replicas may briefly disagree but converge to the same value once writes stop.
Eventual consistency relaxes the guarantee that every read sees the latest write, in exchange for lower latency and higher availability. Given no new updates, all replicas eventually converge. It’s the availability-favoring side of the CAP trade-off.
Worked example: in a cart replicated across regions, a write in one region takes tens of ms to reach another, so a read right after can miss it — but absent new writes, all replicas converge. Gotcha: ‘eventually’ has no fixed bound, so read-your-own-writes needs sticky routing or reading the primary; fine for a like count, dangerous for a balance.