The areas where most improvement can happen are in things that have nothing to do with emulating amps and fx.
- Onboard UI usability. This could be solved by a dedicated controller hardware unit, proper mobile editors, better MIDI knob control support etc.
- Form factor. Even the FM3 is on the big side and for example the Quad Cortex is way more sleek as a design while packing in more I/O and footswitching than the FM3, no matter how you feel about the QC as a product otherwise.
- Use as an audio interface. Still some issues on that front like not supporting anything but 48 KHz sample rate and not having the conveniences of a dedicated audio interface for controlling in/out levels, routing and mixing in the digital realm. I still use a dedicated audio interface for no other reason than because it's a bit more convenient in some areas for everyday use.
- How powerful the cheaper units are. FM3 could be a bit faster but covers the usecases of most users who just need regular fx chains rather than heavily parallel, highest quality this and that.
With cheap capture players like the IK Tonex pedal, and probably a slew of even cheaper products from Asia to come, making amp modeling a commodity to a degree, that definitely puts Fractal into a hard spot as the expensive, pro-grade tool.
While Fractal has its following, there's going to be a lot of "why should I pay $xxx for Fractal gen IV when I can just spend $yyy and get this cheap box that is more than good enough and models every amp ever made if someone has captured it" going around from the more hobbyist crowd.
It will be interesting to see how Fractal navigates that as the hobbyists (including myself) are still a significant portion of the market.