icosphere subdivision levels analyzed: 162 and 642 vertices
seven Heegner seed sites chosen by greedy maximin on the 162-vertex sphere
Heegner labels assigned to those seven geometric sites by minimizing the normalized pairwise-distance mismatch against the exact 7x7 Hecke distance matrix
label assignment mismatch (Frobenius norm on normalized distances): 2.152094
assigned site labels in geometric site order: [7, 19, 3, 67, 43, 11, 163]
162-vertex mesh
first nontrivial bulk-like mode: 0.16311747088501743
wall-gap-state count under the wall-vs-bulk classifier: 0
This spherical construction is deterministic once the maximin site set is fixed, but the embedding of Heegner labels onto the seven geometric sites is still a modeling choice resolved here by best-fit distance matching rather than a theorem.
The 162- and 642-vertex spectra can be computed in full, so there is no sparse-solver ambiguity here.
The useful test is whether low modes become wall-dominant before bulk-dominant. The note reports that classifier directly rather than forcing a continuum interpretation that the discrete sphere does not justify on its own.