Comment 6 for bug 1270382

This seems to be mainly effecting slaves in Rackspace (ironic since that's where the DNS record in question is hosted). My best guess is that their resolvers get overloaded from time to time.

One possible workaround would be to prime a local resolver cache on each slave with multiple retries to look up the names of each of the hosts we need to communicate with during the course of normal job runs.