I will try it - great idea - but that assumes the IP is exclusive to the test site ( which is likely but not guaranteed),
But the thing is that I shouldn't need to do that. Something is fundamentally wrong with the way the DNS and IP addresses are being allocated when using the speedtest credentials. Google indicates a number of other people complaining about the same thing, but it obviously works most of the time for most people otherwise BTW would have noticed by now
At any rate everything has remained 100% perfect since last week, with 75+/17+ speed tests with very few wobbles in the lines in most of the charts, and no noticeable packet loss.
I'm crossing everything I can cross that it will remain this way. I would find it very annoying not to know what the problem was though.