Replication is a fundamental problem of distributed systems. Existing works have focused on strong consistency, where replicas perform updates in the same total order. The downside of this approach is it requires consensus[1]. An alternative approach is to guarantee eventual consistency, which means you update the local state and propagate the updates. On conflicts, you could do merges/roll-backs.[2]. In eventually consistent systems, if there is a network partition or more than half of your replicas are crashed, you could still make progress.[3][4] As we can see, both require consensus to achieve the desired property. (reconciliation in EC requires consensus to ensure that all replicas arbitrate conflicts in the same way.