So I got the accelerometer from the local distributor and run phase calibration and input shaper calibration and observed 0 improvements. Shaper freqs were way off from the factory setting. The fact that there's a resonance at speed that is used by default structural profile doesn't help. If it was 20hz above it might have gone unnoticed because printer would mostly pass through that speed during acceleration or braking phase. Still even when it passes, there's this buzzing sound which is noticeable.

At this point I feel like the situation is worse than with Ender3 where manufacturer provides 0 support, but community knowledge is vast and you can always fix any problem.

/rant

For me it happens on 1 axis move (x louder than y) at speed of 80mm/s and 2 axis diagonal move at 57mm/s. Diagonal move causes 1 belt to move at the same speed as both when doing 1 axis so it adds up.

At that speed I get 253Hz frequency (measured by phone).

When moving gantry/head by hand on printer that's turned off I can feel roughly 6 steps/"clicks" per 1 belt tooth movement. Belt has 2mm tooth pitch so we have 3 clicks per mm. 80×3 gives us 240 which is close to 253hz resonance speed.

I suspect this is resonance happening in the structure that makes such vibrations. Nothing to do with actual motors, pulleys, idlers.

I also tried measuring freq when changing belt tension and it seem unaffected.

Raised it with support and yet to hear back from them.

I have now taken both Y-rods off and on quite a few times. Typically, I remove/replace with only the side panels and top removed (including the top of the metal frame). First loosen the 4 bearing screws. Then, I loosen just the front rod clamp (this is the trickiest part, as the bolt is blind and behind the Z rod), then remove the two screws holding the back side of the Y rod holder to the gantry, then while supporting the front of the rod, slide it backwards. I can then gently take the rear Y rod clamp, Y rod, and bearing away from the printer. I only had to drop it once and clobber my build sheet to figure this out.

I also have noticed that, sometimes, depending on how I assembled things, the resonance moves around. The resonance is not even my main issue, but the rattle/buzz it causes. I've worked with support, and now even prodded around with a mechanics stethoscope, to no avail. I went as far as running my gcode to replicate the issue with the left Y rod in place, but the bearing removed, holding the left gantry up by hand, and the noise was completely abated. With the stethoscope, the noise does seem loudest around the left side for me, and I can also eliminate it by pushing up on the left bearing while it runs, but clearly having tried 4 different bearings, it's not the bearing itself but something else causing it.

The sound is produced both when printing and when moving gantry by hand (motors are not engaged). Initially it was very bad and I followed the advice in this thread to unscrew the linear rail and re-tighten. That reduced the scale of the problem or maybe shifted the resonance frequency.

It is still bad. Here's the example video of it printing with 0.20 STRUCTURAL profile.
https://youtu.be/1ND6Nls4ubY

You can see that when it prints inner perimeter at higher speed the sound is very pronounced, but when it does outer perimeter with lower speed the sound is gone.

I only read about late tightening of Y rod clamps after mounting gantry so didn't try that. Did anyone had success doing it without stripping most of the outer case apart?

Another point is that if I completely loosen the belts vibration seem to go away. Going to raise it with support if I can't get any further.

Posted by: @remix7657

Cool. Thanks for confirming. I’m starting to think that Prusa is crazy for not using thread locker ANYWHERE on this printer. 

Starting to regret following the instructions exactly and not using any for my kit assembly. 

I did apply threadlocker but to be honest, it's mainly for the sheet metal frame. A lot of the internal components use printed parts and I didn't want to risk applying it there.

Posted by: @raaz-2

That's not necessary.

Yep, after multiple weeks, I completely agree. The end-stops are fixed in place by two screws in the CoreXY frame, so they can't be out of square.
And if they are, there are way bigger issues with the printer, than squaring the gantry..

I also move the gantry - center the extruder, move the gantry back a bit, then use a caliper to measure the distance from the brackets to the front to make sure it's still square.  That's slightly better than eyeballing that the two sides hit at the same position, but that level of precision isn't really needed.

 That's a good method! I trusted my eyes and ears probably a bit too much and should've used my caliper.

Starting to regret following the instructions exactly and not using any for my kit assembly.