Gr is pretty Gr, I agree

- I hear the grace notes in m. 4 as actual melody notes in the original instead of an ornament. Thoughts about writing them as 16ths instead?
Writing as 16ths would be too slow, and writing out as 32nds would be less easy to sight-read (as well as take up the limited horizontal space in an already fairly compressed system). I would prefer to keep the current notation, though I do understand your reasoning; I see it as trading-off a small amount of phrasal accuracy for a more streamlined user experience. One benefit of the current notation is that it allows for the notes to be played at the speed able to be executed by the performer, either around 16ths, 32nds (as I prefer), or somewhere in between.

- I think it might be nice in the to clarify where and how the crescendo climaxes at the end through the addition of another dynamic placement in the last measure.
The endpoint of the "cresc." in m.27 is the "mf" at in m.5. If the player overshoots a little from not reading ahead, I feel that would be fine (I actually do that a little myself in the performance demo, crescendoing to forte and backing off slightly at the loop).
A little oddity I noticed - I find it weird that m. 4 has an outgoing bass clef when... there is no clef change between m. 4-5 because of the ossia. Maybe it would be best just to remove it?
But it's just an ossia, right? The clef change makes sense to me since the primary way to play the phrase in this transcription has a bass clef.
One of the problems is that I'm not using ossia in the technically correct way (i.e. as a clarifier of ornamental markings), but rather as a way to write an alternate start to a repeat passage in a register outside the main clef. So, while the ossia notation would seem to imply that it is optional, my intention is for that to be primary execution. I could avoid this ambiguity by not using repeats, but that would increase the page count. I think I'll opt to keep the cautionary clef, if only because eliminating it can only be achieved by disabling cautionary clefs globally in Finale's document options.
- The Layer 2 notes in m24 look very squished vertically. Is there a reason for that?
Finale's default stem lengths and beam angles are very long compared to most publication-quality works. I use the Patterson Beams plug-in to correct this. The flag of the 16th note intersecting with the notehead is actually very common in many fonts outside of Maestro.
- You have nested slurs (i.e. a slur within a slur) on your grace notes in m16, m20, and m24, but not in m12. Intentional or unintentional omission?
Mostly, I couldn't find a good place to put it at m.12. Below, and it would disrupt the readability of the ledger lines; above, and it would be uncomfortably close to the phrase marking. Gould does permit omitting individual grace slurs if they're combined within a standard slur (pg. 130), so that's what I ended up doing:

- Why not have the "con Ped." from the beginning instead of starting at m4?
I assume pedaling by default, and only explicitly write "con ped." if I feel the notation might lead the player to believe otherwise. The way m.1-4 is written, the only way to play it would be with pedal. On second thought, I think I'll omit the "con ped." at m.5-12. I had worried the accompaniment looked "staccato-y" since it was derived from the marimba line. On the plus side, this would help clear up any implication that m.1-4 is "senza pedale" when it's not.