I am only here because the Ookla based tests failed to bring any result.
Mainly because your tests states that I have currently an upload of about 100kbps...
Great ! I have analog modem speed.
Well, I pay for 100Mbit down / 6Mbit up (fastest available in my country [for private users] ... well at some places you can get more down but only with even less upload)
But I also think your testing method is wrong.
It has a systematical error in it.
"Why" you ask.
Ok, you think multithreading is bad because you should be able to max you connection even without. That might be true if there wouldn't be so much servers in between.
And also if you want to use your connection with one device only. (For myself I have normally [right now disabled for testing purposes] 9 devices connected, sometimes even more[not turned on all the time], and even more if I have guests)
A connection to the US goes through a very limited bandwidth. I never get even near 25% of my connection speed no matter what. Some locations are even worse (Africa and Japan .. Japan is a guess). While I get near full speed within most of Europe (Some are better, some are kinda underdeveloped).
So what I really want to know is the speed from my router to the next possible internet server.
And here is where the tests are failing. Even the underdeveloped multithreaded one.
Single thread: There are too hops and borders and stuff in between to make a real statement. It might be ok for 6Mbit lines or less, but higher speeds are definitely tested with the hand breaks on.
Multithread: The first two tests I run with all servers enabled. Result: 6 mbit. It was because all connections where treated as equal and the slowest ruined the average. Next tests was "only" >Amsterdam, NL => eu.testmy.net< and >Cloudflare (Global CDN) => cloud.testmy.net< result: average 92Mbit. And that with my fucked up upload ( current average: 172kbit but the last 4 sucessful tests where between 46kbps and 81kbps and some even timed out... Autotesting every 5 min. I am at #10, but only 5 passed [can see because some of the results recorded are more than 5 min apart])
I am also able to double check the download since there is a bandwidth monitor build in in my router. I can see that I reach the speeds stated by ookla tests since my router gives me the same data. So they can't be as unprecise as you claim them to be.