As I said, USB audio sample data is carried in packets, albeit small, which arrive at a rate defined by the USB specification. This rate is usually 8 kHz. The audio sample clock (44.1 kHz, 48 kHz, etc.) does not define the rate at which these packets arrive. In fact the audio sample clock is not present on the USB physical layer at all, but is an implied property of the audio data which is negotiated between the sender and receiver. There is no per-sample “timing” of the audio data on the USB bus; the receiver has to reconstruct the timing using a sample rate clock running at the negotiated rate.
For example, if the negotiated audio sample rate is 44.1 kHz then the USB packet will usually contain either five or six samples (44.1 / 8 = 5.5125); these arrive at the receiver effectively at the same time (each complete USB packet is normally DMA’d into buffer memory by the USB controller hardware). The receiver then has to arrange for these five or six samples to be clocked into the DAC hardware at the proper rate - that is, 44.1 kHz. Nothing on the USB physical layer maintains the 44.1 kHz / 22.7 us timing between individual samples.