In sharded clusters some "advanced" maintenance operations like manual chunk rebalancing, performed by DBAs, can benefit from stopping the config metadata changes (i.e. balancer, splits).

It should be usable, something like sh.freeze() and sh.unfreeze(). Of course, if somebody calls this, they should be ready to handle the implications, e.g. jumbo chunks.

Currently you can shut down the second or the third config server to achieve the desired effect, but I'm hesitant to call it a solution.