That's interesting because I seem to have observed VM's network/equipment appear to have two operating "modes" - one with low latency, high speeds, and considerable packet loss just as you describe, and the other one with moderate latency, moderate speeds, and zero packet loss.
What I observed yesterday while I think they were doing some capacity upgrade work, was that at one point I reconnected to a different upstream channel to usual. At this point, I got slightly higher speeds than usual, and also slightly lower pings. But packet loss shot up from zero to about 10%, stalled web pages, etc. as you describe.
Then about an hour or so later, my ping increased again, and packet loss dropped to near zero in an instant - while still on the same channel. Clearly, something changed, whether deliberately or not, and not gradually either. And it wasn't due to congestion or high usage.
If you look at the that graph, between about 9:30am and 10:45 packet loss was consistently high, but at 10:45 my ping increases and packet loss almost disappears (save for the further two reboots/loss of service at 11:00/11:20.
Incidentally during the periods of high packet loss, my ping would stay low. Normally when I saturate your upstream doing a big upload, my ping would increase to the 500-1000ms range but still not much loss. Yet in this period my ping would stay at ~30ms even while saturating the upstream, so it looks like something to do with VM's buffering/scheduling parameters.
I can't really explain it, but put simply, something VM did - some setting they may have been tweaking somewhere - managed to cause 10-15% packet loss all of a sudden, in some way that was completely unrelated to load, oversubscription, or line/signal quality. Maybe a QoS or timing parameter, maybe something else. But still, your problems might not be completely capacity related.