USB Audio Class 2 with asynchronous isochronous transfer is clocked by the D/A conversion clock in the DAC. It tells the upstream device to send more or less data based on it’s buffering status. The D/A conversion clock is the only thing that drives this. What ever other clocks are involved in this are asynchronous to the USB clock and are involved in transfer speeds exceeding the audio rates by way more than 10x.
What comes to USB sources and relation to DAC output are unrelated to such other clocks and more related to avoiding analog domain pollution of ground links or voltage rails when the DAC in question doesn’t possess galvanic isolation for it’s USB implementation. You can use USB isolators for such cases to clean up the USB. This has nothing to do with digital side implementation or clocks and everything to do with analog signal design of USB and PCB layout design.
Putting I2S on a cable is bad idea in my opinion, and the existing de-facto “standards” have clock at wrong side of the link. Since it should be at the DAC side, while these “standards” put it on the source side, recreating same problems that we had on S/PDIF and AES/EBU in the past. Requiring PLL clock recovery at the DAC side.