Actually what it does is looks for large changes in the contact normal between evenly spaced sample points, and then subdivides the sample rate until it has two samples 0.002mm apart where there is still a contact normal change, and marks that as a discontinuity.

It is only able to detect concave cusps. However there you can pick up convex (upward) sharp edges (which you don’t want) if you are doing this with a flat bottomed cutter.

Does the pencil-milling path generation rely on the low-level drop-cutter function?

If so, for a single drop-cutter call can you actually make contact with two triangles at once? And detect the cusps where we want pencil-paths this way?

OR, do you call drop cutter at position x and get one contact-normal, then call it again at a slightly moved position x+dx and get another contact-normal? If these two normals differ from each other sufficiently then we have a discontinuity and should generate a pencil-path somewhere between x and x+dx ?
How do you differentiate between convex “hills” and concave “cusps” with this strategy?

thanks,
Anders