Whatever interface one uses they all can be less or more depending on one's optimization. While I2S has some inherent advantages that doesn't mean it will always sound better than usb or whatever. Clocking and power supplies to the clocks is of utmost importance.. Three categories of clocks, femto, txco, oxco, in theory the order of quality least to most is femto, txco, oxco. In reality a well implemented femto clock could outperform a less well implemented oxco clock. The quality of the clock itself and more importantly the quality of the power supply to that clock matters a lot.
The above holds true whether comparing I2S implementations or comparing usb vs I2S implementations. The most optimal implementation for either of these interfaces would be a well implemented oxco clock in both streamer and dac. Good luck trying to find that in either streamers or dacs. space limitations for the power supply alone make it very difficult to implement. Femto clocks are easiest to implement with these space limitations, cost considerations probably play a role as well.
So, wide variability in clocks and interfaces and we haven't even gotten to differences in motherboards, processors, RAM, operating systems, linear power supplies, music player apps. All have a bearing on sound quality.
And then we haven't even gotten to differences in optical vs ethernet, switches vs routers, digital cables, modems, linear power supplies to all these devices, and I may be forgetting some things.
Some believe you can just slap in any streamer and get the same quality of sound with a fully optimized streaming setup. I and many others have experienced otherwise so believe whatever you want to believe.