Turn off balancing
Bring down one config server
Bring that config server back up
Then try to turn balancing back on. The write to the config database will fail with a transport error indicating the mongos can't reach the config server that was just down. If you try the same write again, however, the write goes through b/c the mongos has reconnected with the config server. There should be retry logic in the mongos to handle this so that the mongos will automatically reconnect to the newly restarted config server and try the update again.