This combined V2 manuscript merges and strengthens two previous QTT release papers — "Part 1: Universal Quantum Capacity Laws" (V1.0) and "Part 2: Precision QED and Cosmology Point to a Planck-Quantized Modular Charge, a Universal Ultraviolet Self-Energy Bound, and a Single-Kernel Universality Law" (V1.0) — into a single coherent release under one DOI. Part I (Capacity Laws) establishes regulator-free, parameter-free, or one-calibrant no-retune tests across noise, heat, transport, and spectroscopy. The central exact theorem is the ballistic noise–heat triad identity ∂_T S_V(0)|_eq × κ_th/T = (4π²/3)(k_B³/e²) = 1.349 × 10⁻³⁰ J³ C⁻² K⁻³, independent of channel count, material, geometry, and Planck's constant after cancellation. Four further capacity results — Planckian dissipation as a protocol, unsubtracted spectral closure, the high-frequency Johnson power null, and the single-kernel QED audit — are derived with the same release discipline. Each is paired with binary falsifiers and pre-registered pass conditions. Part II (Precision QED) moves the same finite-capacity logic into precision QED. A QTT capacity principle states that every localized charged excitation is completed by a Planck-bundle modular charge, giving the regulator-free self-energy ceiling U_self(n) ≤ n²αℏc/(2ℓ̃). Precision QED becomes a no-retune audit: after fixing α once from the electron anomaly, the same leptonic spectral supply K(ω) must reproduce windowed Lamb shifts and hyperfine splittings across hydrogen, helium ion, muonium, and ³He⁺. Five benchmark windows pass with diagonal χ²_diag ≈ 0.38 (correlated form χ² ≈ 2.3 for 4 dof; χ²/dof ≈ 0.58) and no coherent cross-window drift. The QTT-novel layer is uniform across both parts: per-address capacity E_cap = ℏc/ℓ̃, the spectral normalization ∫₀^∞ K(ω) dω/(πω) = α, the self-energy ceiling, the real-dial phase generator J (with J² = −1, no primitive complex unit), and a finite ultraviolet ledger that turns formal divergences into capacity statements. The combined paper makes no fit to any of the empirical tables. It is deliberately unforgiving about attack lines: it separates strict parameter-free from one-calibrant from protocol-only, keeps hadronic, weak, recoil, finite-size, Zemach, and polarizability inputs outside the leptonic kernel, distinguishes finite observable closure from a proof that textbook renormalization is unnecessary, and gives binary falsifiers for every QTT-novel claim. Substantive V2 strengthenings beyond simple concatenation: (i) a single executive claim that bridges both parts; (ii) a single deduplicated bibliography with 24 references; (iii) a unified attack-line closure matrix combining both papers' closures; (iv) one cohesive five-tier experimental release program; (v) the single-kernel QED χ² discrepancy between the two source papers is resolved by quoting both the diagonal form (χ²_diag ≈ 0.38) and the correlated form (χ² ≈ 2.3 for 4 dof, χ²/dof ≈ 0.58) consistently and explaining the difference; (vi) explicit cross-references to the broader QTT corpus (framework manuscript, Schrödinger paper, Navier–Stokes paper, Born paper, measurement paper, Einstein–Hilbert paper, Standard Model charge ledger, neutrino mass-squared ratio) situate this release inside the corpus rather than as an island; (vii) the AI use note is updated to the corpus-standard wording. Status: combined release-ready preprint inside the QTT axiomatic framework. The capacity layer claims (Star, StarStar) are explicitly marked as speculative and falsifiable. The established (OK) results are not rebranded. The QTT-native leap is the capacity principle, marked StarStar throughout, with explicit falsifiers in every section.