s1108 — d-index build: ordering bit
  ladder R_d = 1/(φ^{2d}−1): R_1=0.61803 · R_2=0.17082 · R_3=0.05902 · R_4=0.02175 · R_5=0.00820 · R_6=0.00312
  sector data (m₂/m₃): ν=0.17325±0.00233 · ℓ=0.05946±0.00006 · q=0.02234±0.00335

Q1 — data-preferred assignment among injective maps {ν,ℓ,q} → d ∈ {1..6}
    (ν,ℓ,q) = (2, 3, 4)  χ² =     56.92
    (ν,ℓ,q) = (2, 3, 5)  χ² =     74.70
    (ν,ℓ,q) = (2, 3, 6)  χ² =     89.81
    (ν,ℓ,q) = (4, 3, 5)  χ² =   4301.45
  [PASS] printed assignment (2,3,4) is the χ²-minimum — consistent with the ladder, and with SELECTION
    Δχ² to the runner-up: 17.8 — the data pick (2,3,4) unambiguously; that is evidence the LADDER fits, not that d was derived

Q2 — rules on REGISTERED features that reproduce (ν,ℓ,q,CKM) = (2,3,4,1)
  affine rules on ≤2 registered features reproducing (2,3,4): 1
    d = 2 + (1)·[Y_R≠0] + (1)·[coloured]
  [PASS] at least one one-line rule exists (the assignment is not structurally IMPOSSIBLE)
  the 1 rule(s) use REGISTERED labels (Y_R, colour) but the map label→rung is NOT a registered theorem;
  the feature menu itself was chosen post hoc (5 features, ≤2 per rule) ⇒ d(ν)=2 stays SELECTED, not derived
  CKM d=1 needs a separate clause in every rule above (a transfer is not a field) — a second selection.
  VERDICT Q2: NO-GO as a derivation. The assignment is REPRODUCIBLE by 'd = 2 + [Y_R≠0] + [coloured]' and siblings,
  none of which is a registered object. Candidate map named; forcing absent.

Q3 — ★ ORDERING INDEPENDENT OF THE RUNG
  Ladder membership = consecutive generations in GEOMETRIC ratio R_d (gen 3 heaviest by construction).
  IO needs a near-DEGENERATE heavy pair (m₁≈m₂, m₂/m₃ → 1 in ladder labels); NO needs m₂/m₃ < 1. On the ladder m₂/m₃ = R_d:
    d=1: R_d = 0.6180 < 1 ⇒ NO
    d=2: R_d = 0.1708 < 1 ⇒ NO
    d=3: R_d = 0.0590 < 1 ⇒ NO
    d=4: R_d = 0.0217 < 1 ⇒ NO
    d=5: R_d = 0.0082 < 1 ⇒ NO
    d=6: R_d = 0.0031 < 1 ⇒ NO
  [PASS] R_d < 1 for EVERY d ≥ 1 ⇒ the ladder predicts NORMAL ORDERING whatever rung ν occupies
  [PASS] even the shallowest rung R₁ = 0.618 is far below the IO floor ≈ 0.98 — IO is excluded by the ladder at ≥ 10σ-equivalent
  ⇒ The ORDERING bit does NOT need d=2 derived. It needs only: 'the neutrino sector sits on the R_d ladder at some
    rung d ≥ 1'. That premise is WEAKER than the printed d=2 and is exactly what Paper 2 already asserts for all
    sectors. Under it, 'normal ordering assumed' (θ₁₂ᴾᴹᴺˢ, δ_CP premises) becomes 'normal ordering DERIVED-CONDITIONAL
    on the ladder' — the premise moves from an observational input to a structural one already carried in print.
  What the rung STILL fixes: the magnitude (R₂ = 0.1708, −1.0σ vs the m₁=0 floor, s1104) and Σm_ν ≈ 59 meV.

VERDICT
  T1 result: (a) d(ν)=2 NOT derived — reproducible by unregistered feature rules (dense), NO-GO stated; (b) the
  printed assignment is the data-preferred one (χ² unique), i.e. SELECTED; (c) ★ NORMAL ORDERING follows from ladder
  membership alone (R_d < 1 ∀ d ≥ 1) — the keystone bit is derivable at the tier of Paper 2's ladder premise, which
  is STRUCTURAL. Route: θ₁₂ᴾᴹᴺˢ and δ_CP premises 'normal ordering assumed' → 'NO conditional on the R_d ladder'.
  Tier does not rise above the ladder's own (Structural) — but the premise is now internal, not observational.

s1108: ALL CHECKS PASS
