I've received a confirmation from Symas, which is the back_ndb developer th= at replication does not apply to this backend. They explained that by desig= n each slapd instance is intended to connect directly to the NDB cluster by= way of back-ndb. High availability is handled by the cluster.