Comment by aeyes

So if these clusters are using binlog replication, do they just ignore the possibility of lost writes and inconsistent data after a failover?