Thats fine in theory - and with most USB receivers nowadays use asynchronous transfer mode so there is no clock recovery at the USB receiver. There is however a buffer and as far as I know some sort of FIFO process to make sure no bits are dropped. Some better transports even have bitperfect audit capability to make sure there is nothing being lost. The only difference I can see happening with an asynchronous USB interface is that if the USB cable or computer has high latency then a larger buffer might need to be used, which in turn requires more processing power etc. but as far as I know should still allow bitperfect output. At worst if there is a huge latency spike then some bits might be dropped. If you have a wireless card then this might happen about once every minute or so with Windows operating system. Linux and Mac tend to have lower latency but it is still a good idea to disable wireless and other unnecessary processes to avoid other threads competing with the audio playback thread/s.
The point of all these buffers and PLL filters etc is to reduce the jitter in the signal to ensure that a bitperfect accurately clocked signal reaches the DAC. Ideally in theory you could have reclocking inside the DAC as well but this is very uncommon at least in the DAC's I have been looking at and most just use bitperfect filtering and clock recovery, so in the real world the transport does play an influence at least in theory as the transport clock is recovered by the DAC as far as i know. Also in theory, at least with the latest USB transports from XMOS, Audiophilleo etc it should be possibly to use any decent USB cable and computer without running into problems. Users of transports like the Audiophilleo tend to agree that the USB cable and computer software/hardware do not make a very noticeable difference. Experiences with the XMOS USB transports seem to be pretty thin as this is a new design but it is becoming very popular, even HiFace use the XMOS now, and some XMOS transports have been compared favorably to the Audiophillo AP2. So in the ideal world these new generation of USB transports should take whatever the computer/cable throws at them and still produce a high quality bitperfect SPDIF or I2S signal.
Unfortunately though a lot of transports floating around use adaptive or older asyncrhonous USB receivers and drivers which in my own subjective experience with a modified HiFace are not 100% robust and show differences in computer hardware, software and cabling. Similarly some DAC's have excellent digital inputs with or without reclocking and seem indifferent to the transport choice, but most do not seem to match this level of performance. My guess is that the more powerful the USB receiver, the better it can manage larger buffer sizes without breaking a sweat. So if you have a DAC with very high quality and powerful USB input such as Anedio, Antelope, Calyx, PSAudio etc it is not very likely that USB or computer will play a significant influence on the system performance. For less advanced DAC's choosing a modern XMOS or Audiophilleo USB transport should take most of the computer and USB out of the equation.
In the real (or imaginary) world nothing is 100% certain, no system is 100% jitter immune, and there will be specific cases where software/hardware including USB will matter eg with older tech USB receivers and DAC's, very slow, noisy, poorly set-up or old computers or other particular circumstances. I think a lot of the seeming explosion of USB cables on the market is a product of these older technologies still on the market (esp adaptive USB) as well as particular circustances, as well as of course paranoid tweakaholics using imperfect systems. IMO the name calling and stubborn minded opposition between ardent objective and subjective thinkers, or rather beleivers and non-believers (yes science and philosophy tend to overlap and in many cases stand in for one-another) is really not constructive. The fact that many from the objective camp have taken a particular attitude toward the "unscientific" sectors of the Audiophile industry and those who support it doesn't help either (not that this has cropped up recently in this discussion). The difference in philosophy over what constitutes audio performance is 100% a matter of opinion - one side seems to insist (at least in particular cases) that audio performance is a purely objective matter, while the subjectively minded would insist that objective measurements are a sidenote, an alternative measurement rather than an overriding and exclusive measurement. Equally the confusion that results when purely subjective and anecdotal impressions are used to qualify performance continues to create confusion about jitter, electrical noise and its effect on digital to analog converters. In my own experience certain forms of jitter can actually improve the subjective quality of order by creating a darker, warmer and more forgiving audio reproduction. Similarly the designer of Antelope DAC's has said in an interview that certain forms of jitter can improve the subjective audio quality of a DAC by hiding certain electrical shortcomings in DAC chips, or otherwise creating a subjectively " more natural" presentation. The problem occurs when a particular tweak is tested and considered to improve performance and therefore posited to lower jitter even though on paper it can be shown to increase jitter and latency. Another interesting phenomenon is that people with warmer/darker gear tend to chose correspondingly brighter or more analytical source and transport equipment, which equally does not always correlate to lower noise or jitter.
It is only human to consider that the particular point of view one has is valid and all other's are not. I am not saying that anyone in particular is exhibiting this point of view, I am speaking purely in an abstract and theoretical manner. This in a sense is essential to being productive, once one decides on a certain course of action it is productive to follow through with this until the task is complete. In the context of a dialogue or at the stage of evaluating possibile courses of action this is not productive as it may exclude certain possible course of action or inquiry which may be valid or fruitful such as considering component selection or setup. I think in this dialogue it is good that we have people from different perspectives that can contribute to the discussion as long as the door is left open to perspectives outisde if not contradictory to that particular perspective.