Counterexample: defect with no off-diagonal coordinate
Provedgpy_exists_defect_pair_no_offdiagnumber-theory
Over Fin 1, the pair h = 0, d = 2, e = 1 is compatible and not good, yet every coordinate equals the distinguished one, so there is no j different from i and no off-diagonal difference at all. The defect prime 2 comes from the i-coordinate modulus itself. This is a counterexample to the unqualified claim that a defect pair's prime divides some difference h_i - h_j with j different from i, and nothing more. Formalization Note: compatibility and goodness are written out in Mathlib terms.
Preamble
import Mathlib open Classical
Formal statement
theorem gpy_exists_defect_pair_no_offdiag :
∃ h d e : Fin 1 → ℕ, ∃ i : Fin 1,
(∀ i ∈ Finset.univ, ∀ j ∈ Finset.univ,
h i ≡ h j [MOD Nat.gcd (Nat.lcm (d i) (e i)) (Nat.lcm (d j) (e j))]) ∧
¬ (∃ w, (∀ j, Nat.lcm (d j) (e j) ∣ w + h j) ∧
Nat.Coprime (w + h i)
(Finset.lcm Finset.univ (fun j => Nat.lcm (d j) (e j)))) ∧
∀ j : Fin 1, j = i := by sorrySource
GPY defect control: exact characterizations and exact sum identities for the compatible-but-not-good (defect) pairs