That's weird... one connection is almost 20x slower than the other one on Parallels RAS connection

Two virtually identical machines, about three feet apart. Configured identically.

Both computers use the latest edition of Parallels Desktop client, most recent build. They are connecting to the same node on the remote side, which is a third party SaaS. The third party is not cooperative when it comes to hunting down connectivity issues and I have to spoon feed them - force feed them, actually - every scrap of every clue I can dig up to get them to track down problems.

I'm having a problem with the parallels connection dropping for a fraction of a second at random, variable intervals. So far I'm still collecting data on my side to figure out what is going on without much success.

However, today I spotted an interesting effect. These two computers have significantly different connection profiles. Connecting to the same nodes. I only see the network interruptions across the parallels sessions. When I ping that node, both computers are as steady as a quartz watch: I see a trace of jitter from 54-55ms but otherwise nothing happens even when the parallels connection drops.

	ping	RAS Latency	RAS Bandwidth
A	bytes=32 time=54ms TTL=59	5ms	10 Mbps
B	bytes=32 time=54ms TTL=59	97ms	20 Mbps

What might be happening here?