ProxySQL has a more simplified version of this, but it’s transparent to the client. Since it’s constantly monitoring the cluster, it knows what replica lag is at any given time. If you send a request to a read replica, it holds the request until enough time has passed for the replica to have caught up.