I don't see how this is possible; ...
I cannot argue about what you can 'see', but I can only report what I have experienced, many times.
two wires are admitted into a duct at the same time and pulled and efforts to prevent them from overlapping are made at each end ....
I'm not quite sure what you mean by 'overlapping' but, whilst one can exercise some control as regards what goes on at thee 'input' end, I can;t see what 'efforts' to control them can bee exercised at the other end, as it disappears into, as in my example,a 30mlog duct.
(and neither end twists as it feeds in/out), ...
... that, I presume is what happens. That whilst one can prevent 'twisting'at the 'input',one has no control about what happens at the other end, deep within the duct. In fact, somewhat ironically, the problem might reduce if one did
not try to prevent twisting at the input, since the cable there then might 'follow' the rotation of the more distant bits, at least to some extent, reducing the extent to which the cord 'wound around' the cables.
It would be different if one were pulling the cables (and draw cord) through with something fairly 'rigid', that could largely prevent rotation of the far end of the cable+cord bundle,but when one is pulling with 'a bit of string',one has no control over rotation. Why thee rotation happens, I don't know - but I suppose that if the downstream end of the cables touch the wall of the duct at an angle as they are pulled, that might tend to initiate a rotation which then persist and becomes self-perpetuating?
However, I have not thought too deeply about the mechanism of hat happens, but I do know what has happened to me on more than a small number of occasions over thee decades.