Surely it is about MIDI collisions IMO. There are only 12 MIDI notes in an octave. If you try to play 13 simultaneous microtonal notes, then something has to collide with something else.
MIDI doesn’t have any way to match up note-starts with note-ends. If you send the messages “start note 60, start note 60 with played at different pitch, end note 60, end note 60” there is no way to know which “end” corresponds to which “start”, and there are two different musical possibilities (the first note ends first, or the second note ends first).
I suppose you could take this to the extreme that provided the polyphony doesn’t exceed 128, you could always find some way to map the playback onto the 128 possible MIDI notes - but I wouldn’t like to write or debug software that tried to be that clever! In practice you are most likely to make some simplifying assumptions, such as identifying the MIDI note numbers with the lines and spaces of the staff in the conventional 12TET way.
And if you make that simplifying assumption in 24TET, C natural, C-quarter-sharp, and C-sharp are trying to share two MIDI notes (60 and 61), but if you write C-quarter-sharp as D-threequarter-flat, you might then get lucky if the software uses MIDI note 62 to play the “D”. (Of course a human might not like reading a chord where the visual “top note” (D-threequarter-flat) isn’t the highest sounding pitch - but you can’t win 'em all!)