Eventual consistency

In distributed systems, replicas and services disagree temporarily. Eventual consistency means given no new updates, all participants converge — but users may see stale reads, duplicate notifications, or out-of-order side effects unless you design for it.

Trade-off: availability and partition tolerance versus immediate global truth. UX must tolerate lag: progress indicators, refresh, optimistic UI with rollback.

On interviews: sketch read models fed by events; explain how you detect and repair divergence.

Common pitfalls: assuming users never notice lag; no reconciliation job; exposing inconsistent aggregates in one API response without labeling staleness.

Checklist:

• Define acceptable staleness per read path.
• Build reconciliation and monitoring for drift.
• Use version vectors or timestamps where needed.
• Communicate consistency limits in APIs and UI.