When running very small grain sims, with grain separation smaller than 0.001, particle just collapse, not stacking anymore.
it seems that the particle move to far compared to their size, but I could not find a way to solve this , beside reducing the gravity force
the file shows the problem

thanks Martin

increase your max substeps on POP solver

since it's ultimately more about too large forces compared to the scale than the scale itself (until you get to scales that start causing precision errors)
you can do a simple test and increase gravity in your large sim by the same factor and they will collapse too

so it's about the ratio between pscale and speed of the particle, which is what CFL condition uses to choose the number of substeps from the min/max range you just need to allow it to use more substeps if needed

thanks a lot, that solved it!