I'm using CB 4.1.0-5005 on an Amazon AWS linux box.
I think I found the issue though. One of my nodes was approaching disk full, just not one of the nodes reported by the warning message. In fact it was the node I normally use to access the CB Web Console.
I occasionally run adhoc queries with a full production set through the console on that node, and I'm wondering if maybe that is what caused that node to be utilizing much more data that the other nodes in the cluster (which have evenly distributed data).