Hi Chao, I reviewed v3 and also ran a real (non --dry-run) pg_createsubscriber with the same --subscription name for two databases plus distinct --replication-slot names. Both subscriptions were created in their respective databases, the slots were created on the publisher, and subsequent inserts replicated correctly. 040_pg_createsubscriber passes as well.
v3 LGTM and -> Ready for commiter. Thanks, Shihao
