My understanding is that Prusa is buying their gears from bondtech, so there shouldn't be any gear profile issues that first party bondtech extruders don't have. I think you're right about the spacing thing, though. I can even imagine that as the gears turn, there's a point in the transition of each gear tooth where the gear does more to push the gears apart than to advance the second drive gear, which could cause exactly this type of effect.
Walter Hsaio did a series of dual-gear extruder designs, at least one of which used a small shaft, with a separate (knurled) filament drive gear and synchronization spur gear for each. These used the spur gear contact as the pivot, to ensure that the gears always had proper contact (though the angled contact couldn't be good for tooth wear). The knurled gear for filament drive should be maintenance-free, since knurl drives shed any ground plastic much better than hobbed gears, and using them in a dual-drive configuration MIGHT solve the speed-variable steps/mm issue that knurl drives usually have, especially if they're not spring-mounted.
https://www.thingiverse.com/thing:2890048
The earliest dual-drive extruder I saw (I believe it was designed by Aaron Double, who has since pulled all of his old designs from thingiverse) used four gears, two of which were mounted above the ones that actually drove the filament, and were there just to transmit force while allowing the distance between the drive gears to vary. You have to be careful when designing a system like that though, because it's easy for the pivot to introduce a little relative spin to the secondary gear. This is only a problem if the spring tension changes rapidly of course, and if it's doing that, you've got bigger problems than the relative phase of your two drive gears changing.
I did a bunch of testing for this type of noise when we switched to the tatsu 2 design, but couldn't detect any at the time. Of course, if you're using a bowden extruder (which the Bukito did, but the Bukobot did not), the bowden tube is going to be more than enough to damp this type of low-amplitude, high-frequency oscillation.
Another solution I can imagine would be to use a gear with a flexure core that constrains it in rotation while allowing the center to be a little eccentric. I think I've seen wade-style drive gears made this way, but it would be hard to miniaturize to this scale. There are other types of couplers, like an oldham coupler, that could be placed between the two drive gears to fix this, which could probably be made smaller, even if they would have more parts. Actually, any of the solutions that we use for Z-wobble decoupling could probably be adapted to solving this problem (though some are obviously better ideas than others).