Personally? I'd like to see a few things:
Properly Crossfading Modulated Loops so I can sweep the loop location with continuous input without the current horrific habit of bypassing the crossfade and producing a rather non-musical series of loud clicks. The problem is much more noticeable when moving the loop toward the start of the sample (opposite the direction of playback), and it makes one of the more intriguing features of the Prophet X completely useless IMHO. Most of the solutions I can imagine for this would result in briefly delaying the change in loop position and I'd be thrilled to make that tradeoff.
Dynamic Voice Allocation between layers. Instead of having a fixed 4/4 or 8/8 voice allocation between layers, let me use a single voice for a lead or bass part, while using all of the remaining voices for chords. If true dynamic allocation isn't realistic, consider noticing when a layer is in mono mode and automatically balance the voice count as appropriate. True dynamic allocation would have more benefits still, as in stacked situations where one of the two layers is a short percussive attack.
A One-Time Random Mod Source each time a voice is triggered, a unique random value that can be used to ensure that multiple voices aren't identically configured (eg: to allow LFO speeds to vary from voice to voice.) Unlike Noise or Slop, this value would stay the same for the duration of the voice.
MPE Support would be great to see in just about any new synth. Hopefully it becomes standard on future Sequential offerings, and it would be nice to see it make an appearance in the Prophet X as well.