The Uncertified Ellipse: The Joint Budget's Missing Certificate, and the Toll Budget at Stage-IV

Two budgets were charged. The first asked for AF-32's joint (w₀, w_a) distances recomputed at the record's own published ellipse for one named data combination, replacing the declared correlation classes ρ ∈ {0, −0.75} under which the verdicts flipped. The certificate does not exist in owned sources: what the corpus carries for the DESI-class combinations is the centre (−0.83, −0.75) and one-dimensional σ classes — the ellipse's correlation is nowhere stated in any file this line owns. Section 1 prints the audit and the obstruction, names the exact revival condition, and leaves the joint table standing at its declared classes with the flip range as its sensitivity. The second budget is AF-23's, refreshed: the toll forms 2.053528 / 2.055000 need σ(Ω_Λ) ≤ 0.0358% to separate at 2σ (0.0143% at 5σ), and §2 re-reads that requirement against the Stage-IV classes with their conditioning printed — Planck 0.81% (22.6× short), the current record 1.066% (29.8×), the DESI-full class 0.5% (14.0×), Stage-IV w₀w_a-marginalised 0.2% (5.6×), and the Stage-IV ΛCDM-conditional optimistic edge 0.1% (2.8×). The toll falsifier (AF-18's F2) remains armed but unreachable through the Stage-IV clock. Calibration: AF-32's bench is ported verbatim — the closed static table (411/200 = the dress's own identity at 0.000%; every other candidate refusing by 2.6–51%), the w₀ budget (centre-vs-candidates |Δw₀| = 0.17 at 2.13σ current class, 3.78σ Stage-IV), and the declared-class joint table all reproduce before the new blocks are read.