There was some good discussion of this in the past on this forum:
Someone in the thread ran a test compiling the libraries on his own which were native to Apple Silicon. The only improvement was during conversion analysis, and it was a smaller improvement than you might imagine (~15%). I know that is still an improvement, but there are probably other opportunities for improvement that are even larger (for instance, I added improvements in v2.4 that made conversions almost 2x faster than v2.3, and that was just from memory optimization in the code.
That said, I’m of course open to re-exploring, the tricky thing is making the libraries standalone and transportable.
-Nate