This article argues that reliable event-driven systems should be designed for replayability as well as uptime. It introduces the concept of Recovery Contracts, a framework for defining authoritative history, replay boundaries, reconciliation checks, and recovery evidence so teams can rebuild state with confidence after failures.