Yes, thanks both for additional testing. You have re-confirmed to me that 'speedtests' can not be relied on.
What's important (to me), is that I can do something useful like drive my car 70 miles in 1 hour (ie I know I've travelled 70mph) , rather than constant tinkering with multiple unreliable speedo's fitted to the car that all say something different (ie none of them say 70mph).
I'm going to take a back seat now in this car, but still genuinely interested from an academic point of view on how you get on in the front seat tinkering your speedo's.
Thanks for your testing, too.
I do think your analogies are a bit of a bust though, at least based on the testing I've done compared to Alex's.
It seems there is some big difference in single threaded performance to different places that I'm seeing that he isn't with current GW, whereas to some places I don't see a limitation, which obviously shows that at least the connection leaving my house to some point in the world is unconstrained.
So to take a car analogy to the extreme. That's a bit like saying this car will do 70mph* (* 70miles travelled per hour can only be achieved spread across 5 passengers when travelling across destinations which start with a consonant, actual road speed achieved in such situations is 14mph). You can see that depending on your requirements these can be important differences...
I don't think anyone is "relying" on speedtesters. The different speedtester programs and servers seem to offer some insight into differences between connections that you wouldn't get if they all behaved the same and started massive numbers of threads to maximise a connection. You might say the speed testers are somehow not functioning, I'd say they're indicative of a particular performance metric in a particular test setup, and it's unlikely the tester itself is inaccurate per se in what it is measuring. That is unless you are assuming the tester is guaranteed to be telling you how fast your broadband is. It's not, it's telling you how fast you can do a certain thing with your broadband. You then have to relate that back to real world and what that means.
The single threaded difference between what Alex is seeing on the TBB file download and what I'm seeing, while I can achieve line rate single threaded to certain locations, seems to be one of the more significant smoking guns here, given there have been grumblings over single thread issues with Zen's network for quite a while that never really seem to get to resolution. Having to use 40(!) connections to approach line rate to that TBB server is just ludicrous. What's up with that?!