I tend to believe controlled tests
Ah, science. Who doesn't love them some science, right? Can you show me which controllable tests you are referring to, please?
The wiki quote is from this "controllable experiment":
https://www.whirlwindusa.com/tech-articles/opening-pandoras-box
You ready to call this article "science test", but not similar tests from other people? It is LITERALLY one person testing himself with guitar and stating that 10ms of an ECHO is not something you can hear? How is this controllable - or even meaningful - test?? And how did you miss in the "outcomes" section this:
Latencies less than approximately 10 ms to 15 ms are not perceived as echoes with in-ear monitors.
That's right, the part you are quoting is about vocalists, microphones and headphones. And the whole article is pretty much about everything else but guitar.
It would be awesome if at some point we would stop using this nonsense article as "scientific argument".
Do a simple test in a very controlled environment. Set up electronic drums to 15ms and ask you drummer to play it against a track with no delay. Measure the latency between him starting play and sticks flying towards you. It will be very small, almost immeasurable, but I insist you not to trust me on that and see (feel) it yourself.
It is not to say you can't play with the delay/latency, you can and you do (all of us do). But stating that some latency is something people can't feel is top notch deluxe premium bullshit. Some might not, we all are different in the end.
Now jokes aside, are we really comparing Vai's and many other's experience with the latency to some bullshit wiki article referring to a random blog post as scientific study?!
I'm not saying he's necessarily right because he is he. But that nerd is 10 times more scientific and experienced than wiki authors.