Looks like the 256 and 512 got reversed in title of the thread. I assume you meant 256 stereo channels which would be 512 audio channels.
As you can see from the image on the top, the [Insert] Hauptwerk button is DISABLED after you've added 256 audio elements regardless of whether they are stereo or mono. This means that you can use up to 256 mono channels (or 256*2 channels if you want to work in stereo).
The dynamic allocation of audio channels is something I've thought about many times. It is good to see you are actually doing it. I would enjoy hearing more about this.
Let's start with the fact that Hauptwerk's default channels allocation is very effective at distributing sounds equally everywhere. Statistically it should work well.
I can not use the default Hauptwerk allocation if I also want to use additional voicing tools outside Hauptwerk (=within the DAW). For this reason the 12 semitone pattern ("Tone matching mode 1" with 12 channels) is the most effective to isolate the octaves to be treated with the DSP audio plugins. Once you have done this you will have in the DAW the mirror of the audio groups you define in Hauptwerk. There may also be a hybrid configuration: it is you who decide how to structure things keeping in mind that the same audio channel can belong to only one group in Hauptwerk.
Similarly in the DAW you will have to decide how to organize all the channels coming from Hauptwerk and going to the speakers.
It is not that one setting is better than the other, it is always a compromise and a choice based on some priorities.
In this case, using dynamic allocation allows you to get considerable savings on audio resources in the sense that if you do not need to play with clusters and you just need a few triads to accompany the chorus then instead of taking 12 speakers (with a static allocation of 12 to 12) you could only take 4 speakers (with a dynamic allocation of 12 to 4) having approximately the same result as 12 speakers. No magic, only compromise more or less effective.
At the implementation level with Ardour DAW you can define Lua DSP scripts and build your routing logic. For example, I also use the ARDOUR.DSP.compute_peak() method to track the status of the 12 input channels in order to make mixing output decisions.
I repeat that this is only a way of doing, the priority is the additional DAW sounds voicing tools, avoid mixing intermodulation distortion and the overall price/performance ratio.