They share the Ground with die USB Port anyways so Common mode Noise over the USB port is unaffected no matter if they use a spearate PSU or not.
So they need to (or better should have) implemented a Common mode filter anyways.
(something sadly not tested btw.)
So real life performance on a custom PC that has never been tested for EMC compliance can be significantly worse.
The separate power supply only helps with differential mode noise. Differential mode noise is no problem too dale with on DC lines.
Especially since they would have use local DCDC converters and Linear regulators anyways.
Sure, in theory yes, and I have no doubt you know your theory better than me. Now I tend to see things a bit more high level. The science behind perturbations is a bit ****** to study because perturbations is just that: Shit happens.
The basics of this is, when problems occurs, it's often not just one thing, here's the example, you have a significant error rate at the receiver:, start with a poor USB supply, add interferences on the cable, add less than Ideal filtering... No problem to deal with? Most of the time, yes ok, but we are engineerering toward not most of the time, but all the time.
Bottom line, it's never a bad idea to discard a potential problem away, even if it's not the main one, even if there are workarounds. Even if the risk of happening is small.
The thing is. Nobody would design a PCB with High power traces right next and parallel to a data trace right? You learn that quite early. For some reason it's OK with USB, yes there are standards and good practices to remediate that, but again, Shit happens.