Look at this result:
Doesn’t it appear as if it only took the surface normal of the initial control point into account, and just moved all other control points according to that normal, instead of the normal at each control point?
Is what I’m seeing a bug or what is the explanation for this?