I think the poster of that claim is confusing "plain" logical replication and multi-master replication. There's a lot of infrastructure in common between both, but the conflict resolution part of async mm didn't make it it into postgres 10...