And another experiment, one-cloud twister. Veeeery slow, so not my favorite experiment. A normal scene with some ordinary not too high cloud this size would maybe take 15minutes, this took almost 2 hours. I guess complicated cloud makes it hard to calculate, and it's not even v3. I would have thought once the cloud is shaped, calculation would be sort of the same as a 'standard' cloud of same height/thickness, but I guess it works differently.
It's far from perfect by the way, just a proof of principle. And the twister more looks like a sudden enormous downpour because of the added fractal warp, which works in vertical way only due to its place in the node series. I think there must be another way to get a little XYZ variation in that thing.