Does this check happen before the actual collapse happens with the "best position" for the edge? If so, are we essentially doing a theoretical calculation about what would happen if a vertex is placed in a certain position?
I'm having trouble imagining a geometry in which this would happen. It seems like a point which "flips" triangles would generally have a higher quadric error than one closer to the middle of the original edge. Is there an intuitive example?
Is this theoretically a problem in the general case, or just when doing Quadric Simplification? If only this case, what's different?
What are the algorithms to detect the "flips"?
Logically, should this be checked in the flipping local operation or the global operation of simplification?