It's the wrapping of it around a ferrite core that does it. Surprisingly little can make a difference at certain clock rates. If you take a cat5 cable running 1G and loop it a dozen times in a 6" circle and put a metal screwdriver perpendicular to the loop in its center, you will see a spike in bit errors. Weaving one wire with another isn't going to be as strong of an effect, and it's fiendishly complicated in terms of an e/m problem, but it could easily be significant.