I think it might be because of the little vibrations in the instruments sounds that we normally don't notice because when played at regular speeds they sound more like one whole note instead of a vibrating one. The low to medium-low notes of the piano for example, or the broken up sound of a shaker. At speed they sound fine, but when slowed down they wobble and scratch a bit more as our ears have more time to pick out individual bits of what's really going on.

Also, the dissonant notes and runs within the chords pop out a ton more in the slowed version because they're given more time to their selves. Normally dissonant or jazzy notes don't sound too bad to us because they're short and eventually come to rest on a more pleasing chord. But in this, they're all stretched out so it comes off more harsh as you said, and yields a more dark, dissonant sound (which is why you see such dissonant notes drawn out in a lot of dark classical orchestra pieces).

Just my thoughts. I could way totally be wrong.

Because it's probably playing the same chunks of sound six times in a row, thus blurring the stop and restart of the same sequence into a shimmering stutter. Those affects are not necessarily inherent in the performance.

The most common way to stretch out digital sound (such as an MP3 file) is to play it back with repeated sample plays. Imagine a song file were like a film, showing a discrete frame of music. Normally an MP3 player will simply grab the frame (or really grab the parts that make up a frame), play it once and move to the next frame. In this case, the frame is going on your aural "screen" six times, then do the same to the next frame.

This is unlike playing an analog tape, where you would simply slow down the tape and get lower frequencies. Instead you get the same notes over and over at the normal speed. In turn you're getting the sound waves wherever they started and stopped in each sample.

If I were digitally recording a spoken sentence, such as "Hey, welcome to the Internet", I would wind up with things like an aspiration (the "h" in "Hey"), a vowel transition (from "a" to the trailing "ee" in "Hey"), and so on. If I play that back with sample repetition, I get a lot of choppy "huh-huh-huh... ay-ay-ay... ayee-ayee-ayee...".

The tool "mpg321" has a simple flag for changing the sample repetition. Thus you can try this with any favorite song and have it suddenly turn into Rhys Chatham.