Steps to reproduce:
1) set up load balanced situation with at last 2 nodes
2) start distributor
3) use the controller port to change the logging level to
off (ie. "loglevel off")
4) shutdown one of the nodes
5) note that any request that gets switched over to the
node that is down times out
6) use the controller port to change the logging level
back to all (ie. "loglevel all")
7) note that requests are now failed over to the single
node, and eventually, the node is tested and taken out
of rotation.