That's my though as well Wolfie, I wouldn't be surprised if its something really stupid inside a router or switch somewhere where it is not hitting the fallback or primary IP number, causes the bottleneck, causes the crashes. I wouldn't be shocked if they have 2 people only working on this, checking each router or switch config file line by line to make sure it all routes right. Then they would have the software guys checking there stuff, problem is they have not been entirely descriptive. It also could be something as simple as opps we miscounted bring up 5 more instance servers boom problem fixed.
