Replicating into cluster is slow

That graph is from the async master correct? It is receiving replication events from PXC? And you confirm you are running with ROW-based replication?