At a guess, the whole business of spinning up a new instance of the sound generator for each new note that's played, (or running a fixed number of instances, and dynamically allocating new notes whichever instance will be least disruptive, aka "note stealing"), and the other business of doing modulation (such as vibrato) that needs to be applied consistently across all instances of the sound generator... remains to be implemented.
In other words, it looks to me like Faust's focus at the moment is on describing the processing in the audio signal path. Polysynths require a whole bunch of cross-cutting management functions, outside of (and between) the (multiple) audio signal paths.