alphabetic order

Ah, because of the CELL_SIZE change below this is changed from 1 to 4.

That's not good, because it invalidates every number in every random map script!

I think the idea behind that 0.5 is to center the entity on the tile coordinate cx, cz.
The distance argument should be the radius, not the diameter, so I wouldn't really agree with the + to * change.

I'm dubious about doing the CELL_SIZE division in this place as well as the comparison with 1 here.
The number 1 seems to mean the footprint size in the rmgen coordinate grid (which is the tile grid, i.e. the numbers in map_sizes.json).
The rmgen library should use only one coordinate system where possible, so why would we do the CELL_SIZE division here only to test against 1 and nowhere else?

I suspect this is the part that fixes the reported "small stone mine inside a big stone mine bug". But this is due to the fact that the footprint of the mines isn't taken into account to begin with and just hapens to be solved with 4 (cellsize).

Given that we only seem to have issues with the stone mine use-case, just adding these arguments at the end of the constructor seems to work, even though avoidSelf seems to become unneeded / redundant with the distances.

Adding these new min/max distance arguments seems like the only way to address the issue if we want to keep using a SimpleGroup where each Object is able to avoid the other objects within the group.

(I wonder if this the right place for maxFailCount. I doubt the callers to place are happy with having it in place, not the constructor. But same is true for the SimpleObject)

(This while loop was copied from the SimpleObject and has the same issue that it needlessly nests if-else statements instead of just using &&, which allows removal of the fail variable)

This loop is actually correct, because it considers all SimpleObjects placed by the SimpleGroup and every Entity placed by the SimpleObject, i.e. everything placed.

I appreciate that there are distances here instead of the magic 1. I think the meaning of avoidSelf is redundant with the distances though.

This condition is always satisfied (usual issues with copying code. Perhaps we can unify those now 10 lines)

(Here the tiles with the placed objects are marked. A thought would be to use the avoidClasses mechanism to avoid resource collisions. But that conflicts with the fact that first all locations should be found, then all or none being placed.)