When memcached is trying to delete a bucket it needs to disconnect all of the clients bound to that bucket. All connections which cannot be disconnected immediately is logged every time we inspect it (which happens every second). This may cause bursts of logging messages from memcached, and in worst case fill up the entire log (if something is stuck / broken).
The current logging is possibly too aggressive and could be reduced. We cannot completely mute the logging (as we want to know if things are stuck or not).