You can do “REMOVE LEAF … FORCE”, if you don’t mind some temporary downtime (until you run ADD LEAF). The “FORCE” will cause REMOVE LEAF not to bother rebalancing data. ADD LEAF will re-attach any data back into the cluster and put you back into the same state you were in before the add/remove.
Yep, your right. REMOVE … FORCE still triggers a failover to keep the cluster online. So when you ADD LEAF it will attach back the slave partitions that were on the leaf before the REMOVE, but the “old masters” need to be copied again (and promoted back to masters). So REMOVE… FORCE saves you some rebalance work, but not all of it. I’m checking if we have a feature request for this.