Yeah, I thought so... trying too hard
Maybe the below will make more sense... choose to read it, or ignore it....
The ground isolation was introduced due to the neglect that, in true honesty -> you are encouraging here, by saying that "The USB module doesn't give a crap about the noise so long as the signal is far enough above that it can still decode the signal".
This necessity of ground isolation creates a monster - a very small & isolated USB module ground fill that is highly susceptible to noise - it virtually can not handle the noise - it can not get rid of it... A very large number of LDOs (voltage regulators, for each IC Vcc pin - best in the industry!!!) is introduced on May's USB module. These LDOs work to a degree... they try to isolate intermodulation/noise
between various ICs placed on that USB module (segregating the noise between ICs), but
they are (LDOs) NOT EFFICIENT in coupling the circulating noise from that small ground fill to the Vbus common. This noise can not be coupled to May's main board common due to isolation (galvanic isolation, which is ground fill isolation between USB ground and May DAC ground)
The only way the USB ground fill can handle the noise is if it is drained from that ground fill. The only path is via a very low noise and low impedance Vbus circuit. USB cable is an integral part of that circuit; it is an extension of that USB module ground fill. Hence, it has to have a very low impedance path between the Vbus source (steamer) and the May DAC module source of the noise. The noise is drained via a USB cable shield, but also via a Vbus cable -
due to a very low noise/impedance Vbus at the source/streamer - it's like the Vbus and the ground at the streamer-end are sitting at the same virtual potential for AC (HF noise). Now, remember your statement: "The USB module doesn't give a crap about the noise so long as the signal is far enough above that it can still decode the signal".
The noise circulating that USB card can be ISOLATED from entering the May DAC main board; HOWEVER, it will still cause havoc to the sound being processed on that USB module itself if care is not taken. This is where we disagree - you think that the noise is not important here.
The designers did their bit - a large number of LDOs (which still can not couple the noise to Vbus common point of return). The user has to have a good understanding of the above and take it from LDOs a few steps further: use very low noise (usually linear) power supplies to power the streamer, use good streamers that have a separate voltage regulator to generate low noise/impedance Vbus for USB card, and use high-quality USB cables. The focus here with USB cables is not on differential pair USB + and - digital signal.... the focus is on the amount of shield and Vbus cable thickness/quality (+ the soldering of those large shield & Vbus cables to those little USB connector pins).