The theory behind it is interesting, but in practice I found it underwhelming for practical applications. The point of using it was to get a performance boost by separating the geometry composition from the node mapping. Unfortunately, even on simple models it produces erroneous protrusions until the resolution is turned up high enough. This slowed things down to the point that it was no longer any faster than OpenSCAD and since it doesn't have STL import functionality, there was no reason to keep using it.

At least it can make circles that are round instead of polygons, so that's a major upgrade already.

I think you can set the resolution with something like $fn=100; if I remember correctly.

Polygon with large but finite number of sides is still a polygon.

When the desired output format is .stl, that's not a meaningful limitation.

But when the desired format is STEP, then it suddenly is a meaningful distinction.