Alrighty, dealing with the perennial FLIP volume loss issue. Simple scene, just a rotating bottle. No volume loss at first, but once it turns sidewise and back it collapses. I've got the grid scale low and particle separation high (which stabilizes it at the beginning). This video is particle separation of 1.5 and 2/4 substeps, although lowering separation to 1 and substeps as high as 4/8 makes no difference, just takes longer to calculate. Reseeding is off, tried it with it oversampled which didn't help. Also tried as narrow band, same result. What am I missing?

- there is an article on CG Wiki about volume loss
https://www.tokeru.com/cgwiki/HoudiniDops.html#stopping_volume_loss [www.tokeru.com]

Yeah, I'd already gone through those. Grid size is low, particle radius high, played with a range of particle separations. The collision is a volume collider that is looking good, and the particle counts at the beginning and end are the same, so it's not particle loss.

Thanks to advice on another forum, I found the solution: Use MPM.

I actually did a very similar project and run into the same issue. I remember raising sub steps (up to 50) did help. From what I've read MPM solver is probably not the right choice for such a simulation.