So, there are two things currently going on with the Debussy example.
The first is, yes, some of the notes from Debussy are actually poorly playable on the piano. One is this: https://i.imgur.com/K0Wuu3Q.png. This is a visualization I built (and which can't be ported to web easily, so unfortunately it's just something I have for now). It's like a scrolling player piano with rising notes, and the numbers next to the bars are what the numerators are. Note there are two 7s in the middle, and pianos are not good at 7s. I've verified these numbers are correct manually, so it's supposed to sound like that and the piano has an inaccurate approximation. This corresponds to this location in the score: https://i.imgur.com/24XY8vR.png
The second is, I suspect there is a mistake near the end. This visualization overlaps the original with the twelve-tone version: https://i.imgur.com/tS7WzRg.png. Note how some of the bars are separated from each other. I'm currently in the process of checking these notes. My original 12ET -> Just Intonation conversion was automatic and I fixed up all the errors I could find, but I could have missed one. EDIT: yes, there was. I fixed it now.
The third doesn't happen in this example, but sometimes composers want to deliberately round, such as in the Circle of Fifths. That would create unpleasant commas. I'm kind of lucky that it didn't happen in the demo I chose. The Circle of Fifths is not actually an issue; first, you can't keep old notes persistent across the whole circle; they will interfere. And second, thanks to Diana Deutsch's experiments on pitch memory, we know that if there are more than 16 notes between the old and new note, you shouldn't be able to discern the comma. However, although the Circle is not a big deal, other rounding intervals can become an issue, like 63/64, where the composer would like the rounding to occur somewhere other than the exact pitch repeat. Listeners are effective at detecting when an exact pitch repeat is slightly inexact, while they have a harder time discerning slightly-off harmonies.
The first is, yes, some of the notes from Debussy are actually poorly playable on the piano. One is this: https://i.imgur.com/K0Wuu3Q.png. This is a visualization I built (and which can't be ported to web easily, so unfortunately it's just something I have for now). It's like a scrolling player piano with rising notes, and the numbers next to the bars are what the numerators are. Note there are two 7s in the middle, and pianos are not good at 7s. I've verified these numbers are correct manually, so it's supposed to sound like that and the piano has an inaccurate approximation. This corresponds to this location in the score: https://i.imgur.com/24XY8vR.png
The second is, I suspect there is a mistake near the end. This visualization overlaps the original with the twelve-tone version: https://i.imgur.com/tS7WzRg.png. Note how some of the bars are separated from each other. I'm currently in the process of checking these notes. My original 12ET -> Just Intonation conversion was automatic and I fixed up all the errors I could find, but I could have missed one. EDIT: yes, there was. I fixed it now.
The third doesn't happen in this example, but sometimes composers want to deliberately round, such as in the Circle of Fifths. That would create unpleasant commas. I'm kind of lucky that it didn't happen in the demo I chose. The Circle of Fifths is not actually an issue; first, you can't keep old notes persistent across the whole circle; they will interfere. And second, thanks to Diana Deutsch's experiments on pitch memory, we know that if there are more than 16 notes between the old and new note, you shouldn't be able to discern the comma. However, although the Circle is not a big deal, other rounding intervals can become an issue, like 63/64, where the composer would like the rounding to occur somewhere other than the exact pitch repeat. Listeners are effective at detecting when an exact pitch repeat is slightly inexact, while they have a harder time discerning slightly-off harmonies.