The important thing about bits is not that they are bits, it is that when they are mis-read, they can be corrected. Can be, not must be, or always are.
The internet does provide a protocol, Transport Control Protocol or TCP, which guarantees that a file can be (eventually) delivered in perfect condition. But the internet also provides another protocol, User Datagram Protocol, which does not guarantee completeness, accuracy or delivery but does try to get the bits out in time. While Qobuz, for example, say they use TCP, they start playback before the stream has finished using their own proprietary protocol. How do they know which packets have errors, or have been dropped completely, and how do they ask for re-transmission? How often can they do this before the stream stops? Same questions for Roon.
Next, USB. USB does not guarantee bit-perfect transmission WHEN USED FOR STREAMING. When set for data transfer, it does not guarantee timing.
Ethernet - same story. Ethernet on its own does not guarantee packet delivery, nor accuracy, nor timing. Fortunately, it is usually fast enough for higher level protocols to make up these deficiencies, but what protocols do Qobuz, Roon etc use on top of Ethernet? Seems to me these protocols are kept very proprietary!
I2S? Never designed for data transmission beyond two chips on a board, I2S has no error detection built in, let alone error correction. At least it has timing.
Bear in mind that the bits streaming from a spinning CD need about a 2k buffer just to detect and correct errors. By design, 4,000 consecutive wrong bits can be corrected - an astonishing feat for 1982 technology. This is before the stream emerges as PCM.
Of course, if the streamed bits can be captured to a computer file, that file can be accurately compared to any other file purporting to contain the same bits. If they match perfectly, then I accept that bits are bits in that instance.