I'll try running it without the vhr setup tomorrow... at least part of the problem does seem to be vhr related; not sure if the slave backends reconnected to the db ok -- both times I killed the slaves first -- but for sure nothing got served after the db went down until I restarted the master. Not sure why the one would hose the other but both times the master server has gone down and stayed down until I killed it. So maybe I didn't need to restart the slaves at all but I won't know until it happens again. :/