$500 USB cable

If you hear a difference between USB cables, it's likely that your DAC is not using asynchronous USB but rather synchronous USB. The latter is much more susceptible to cable quality than the former. Because in synchronous USB the jitter contributed by the cable can affect sound quality. With asynchronous USB there should be no difference in sound quality between USB cables assuming the receiving DAC does not rely on the USB power which any reasonable DAC won't.  

I have a degree in computer science have some knowledge of how data is transmitted between devices. The data is split up in packages of bytes. Then the procedure is basically the same as when you mail a letter at the post office. There is no data loss since if the package is verified by checksum when received and if not correct requested again. Since the bandwidth capabilities of the cable is many times that of our (audio) stream a buffer makes sure our bits can be aligned in correct order even if we would need to request a package again. A 3 minute song in uncompressed CD quality is about 31 MB. An old USB 2.0 cable can transfer 35 MB/s and would be able to transfer the whole song in less than one second (assuming that the dac has large enough buffer). USB 3.0 specced cable can transfer 625 MB/s (5Gbit/s) the whole song could transferred in 1/200 th path of a second. The rather old tech USB 2 can even transfer the same 3 minute song of uncompressed 192/24 in less than 6 seconds. What seems to confuse some people is that a CD transport might not be able to read a disc that migh have fingerprints or scratches and tiny read errors might end up as what we know as jitter. Much of this is solved by buffering that will allow the CD transport to try again to read. Many modrn dacs also have the ability to estimate the data and smooth out possible errors . This is however not the case with files on you computer. There are no read errors. The data is perfectly packaged and shipped to you dac.

A little addition about synchronous data transfer. Assuming there is no buffer implemented in the dac (very unlikely) and all packets sent in correct (synchronous) order it would take 200 failed attempts to transfer a package before it would affect the sound stream.

People assign analog attributes to digital cables. Warmer, colder, brighter etc. I just wonder which bits are added, changed, or entirely missed when the sound turns warmer or brighter etc.? If bits are missed or added it will likely cause nasty noises (if not corrected by the error correction mechanism of the receiving end).

There is no error correction in S/Pdif transfer, but that is not the point.
S/Pdif is coded such way that word clock for D/A converter can be extracted from it (from transitions) to keep CDP and DAC synchronized. This clock has to be rock solid (no time jitter) otherwise additional sounds will be added on analog side. For instance, if you play pure 1000Hz tone and have a lot of 60Hz noise then your S/Pdif stream of 1s’ and 0’s might have 60Hz jitter (vibrate in time back and forth 60 times a second). It will produce on analog side pure 1000Hz tone plus (mainly) 940Hz and 1060Hz tones (sidebands). These tones will be at very low level but still audible, since not harmonically related to the root frequency of 1000Hz. Amplitude of these extra tones will be proportional to the amplitude of the vibration (jitter). Since music has a lot of tones there will be a lot of additional tones added that will be perceived as noise or lack of clarity. There is a lot of sources of jitter, digital cable being one of them. It might add jitter because of reflections in the cable affecting/deforming edge of transition or it might be susceptible to ambient electrical noise, that when added to signal affects recognition (threshold) point hence affecting moment when transition is recognized. Jitter might also be caused by noise added to signal in CDP or DAC itself. Since different digital cables affect jitter differently (one has better shielding while the other less reflections) there will be a difference in sound.

I have a hard time picturing an optical transmission of very slow data stream for 50 centimeters being much wrong.  A single optical cable of similar diameter is able to stream data for cities of millions for kilometers.  Do you have any data or measurements that suggest what you're saying is correct?