Literally the current top comment for this article explains how WebGL values are used to more accurately determine which video stream is best to send to the video player, since other methods can return inaccurate or insufficient information. Because you want to make sure the video format has hardware acceleration for decoding, to not destroy a mobile user's battery.

So that's how they're related.

And let's not assume guilt and then require proving innocence? It works the other way around. It's the burden of the person making the accusation to form a strong case. Which is not done here at all.

Oh, I missed that comment---I thought the top comment was same at that time, but it has changed since then. That suffices as "explicit alternative explanations". Thank you for pointing it out.

I still believe that the presumption of innocence should not apply in this case, because fingerprinting is already close to guilty (if you don't agree to this premise there is no point for further discussion, so please refrain from arguing against this specific point). You are right that feature detection is not fingerprinting but they are virtually indistinguishable; given how widespread fingerprinting has been, it was enough to suspect so.