Sure you don't have that backwards? USB 2+ has isochronous transfers with guaranteed bandwidth and latency that come with guaranteed start-of-frame timing strobes every 0.125ms with a jitter spec much better than that. Ethernet... doesn't.
It depends on the type of USB audio device. USB MIDI devices, for example, use bulk transfers because they need reliable communications to handle things like sysex sends. Additionally, Ethernet, through the AVB suite [1] does have time reservation, synchronization, and other things that are very useful in dealing with realtime data.