As xnor says, nobody is really worried about data loss over USB.
What seems to cause the major concern is jitter.
The USB audio interface (the one that everybody uses, not the recent one) seems tailor-made to excite concern over jitter, or perhaps it's just that in the absence of any seeming concern over data loss, audiophile paranoia had to search out another cause for concern. Certainly the interface was designed without any awareness of the extreme stance likely to be taken by some audio enthusiasts and professionals (salesmen, not excluding sales engineers) and hence it is extremely awkward under conventional USB Audio Class 1 using Microsoft's driver to provide a stable, low-jitter clock at the player, which would have relieved any concerns about jitter. The clock must be recovered from the transmitted data.
The problem is that the player device has no control over the rate at which data is transmitted. This means that if a local clock is used, eventually a buffer over- or underrun will occur. This is addressed in the latest USB standard, but the latest standard is not supported by Microsoft, so that although USB Audio Class 2 has native drivers for Linux and MAC OS's, Class 2 Audio DACs remain (prohibitively IMO) expensive (especially since nobody has satisfactorily demonstrated that jitter does seriously affect devices using the previous standard).
It's a typical audio traffic accident, with all the usual suspects conspiring to ensure that the maximum fear, uncertainty and doubt are excited amongst customers. Good for business.