As I posted earlier, a simple tracert shows that the connection is being lost somewhere between an NTT system and a verio system. Give it a try using Stanford's tracert tool online. Also, Verio IP backbone routes have been known to be down for weeks, not hours or days.
I'm not a network specialist, and the problem doesn't necessarily lie within the parameters above. But the problem does appear to be at the network level.
The trace-route shows the packet loss is occurring before it reaches the host. That's the entire point of my previous comments.