Comment 2 for bug 1272746

Seems we have a resurgence of this with a vengeance. At the moment, we're getting a significant number of icmp/net-unreach responses for all regions in hpcloud trying to talk to pypi.openstack.org (and no sightings of similar issues between instances/tenants/regions within rackspace). This could imply network trouble in hpcloud, at rackspace's border, or on the backbone somewhere between them.