In our case async replication was acceptable, so we had a failover solution available.