In theory quantisation should snap everything into perfect time, and quantised swing will still be perfectly mechanical. But there are a number of reasons why the timing might drift.

This is trivially easy to test. Use the pads to record a groove with a standard collection of samples. Replace the samples with clicks. Patch each track to the separate outputs and record into a DAW. Any timing shifts will be very obvious, down to individual samples.

With respect to Mr Linn, he's quite wrong about randomisation being uninteresting. Real drummers play with a Gaussian distribution, even when they're playing to a click. The tiny variations add a surprising amount of interest and liveliness.

But this works best as a dynamic performance effect, and I've never seen it implemented in a drum machine. Baking a randomised uniform distribution into a short repeating pattern is a lot less interesting.