The main feature which was added in 2.1 was the replication scheduler, which makes it possible to run millions of replications concurrently. That makes the data

The main feature which was added in 2.1 was the replication scheduler, which makes it possible to run millions of replications concurrently. That makes the database-per-user model much more feasible. As for throttling, the CouchDB replicator gracefully handles HTTP 429, but that wouldn't help with PouchDB-managed replications. I can't really speak on PouchDB, but maybe it's possible to limit the replication on the client side, since that's where it's managed?