Indeed, Replicache works this way, using server reconciliation (one part of client-side prediction): https://doc.replicache.dev/concepts/how-it-works