This paper is the forty-second in a sequence developing an interpretive framework that redescribes physical reality as a causally consistent history of information updates rather than as a collection of independently existing objects. It is the recovery panel of the twelfth grouping — the U3 construction program — and it adopts nothing new. Paper 41 adopted, at the claim level L3-A defined there, the common record-history kernel Pₖer (one classical-quantum GKSL-type generator whose fixed Lindblad operators are export-event operators driven by a classical, adapted record-export intensity), computed a nonempty admissible coefficient space, and left four obligations to this paper: the recovery conditions RC1/RC2a/RC2b and the single calibration RC3, under a binding ledger (one calibration, spent on lambdadec only; lambdadiff carried as a gate-bounded prediction band; thresholds never re-tuned). This paper executes them. RC1 is proved as a theorem: in any sector containing no certified export channels, the kernel's matter/order dynamics is exactly unitary — a sector statement, not a coefficient limit, and not a claim that unexported mass ceases to gravitate. RC2a is proved as a bounded-residual theorem: the Newtonian baseline drift is fixed by definition to the imported Newtonian classical-quantum mean dynamics, the export correction is shown to be cell-mean-zero by the decomposition bookkeeping (with the adopted sourcing law's first moment checked — a real failure mode, not realized), and the residual mean-source shift is dipole-suppressed below every declared class tolerance deltag (class) ; the honest status row records that no current class co-locates audited export density with a Newtonian-verification measurement, so the world-side probe is today bounded, not probed. RC2b is split in advance: an executed compatibility check — the Jacobson-route hypotheses of Paper 36 reduced row by row to declared criteria with declared tolerances, of which the normalization row is satisfied (G an input, never an output), the entropy-monotonicity row is satisfied under a displayed record-basis pinching condition, and the local-equilibrium and Clausius-defect rows are fixed as criteria whose concrete exhibition and evaluation are assigned to Paper 43's audit program (no row failed; the which-entropy circularity imported as a live defect) — and a typed covariant debt (relativistic microcausality, dressed sources, species), named and not discharged. RC3 is spent exactly once, on the single calibration benchmark declared before any dephasing datum was consumed: the measured measurement-induced dephasing of a dispersively read superconducting qubit (Schuster et al. 2005; channel model Gambetta et al. 2006). The audit is non-vacuous by construction: certified exports are the output-port photons insofar as their field modes are amplified into the retained quadrature record at the declared grain — the certified intensity is nuₑff = etacert kappa nbar with etacert = etaₑxt etaᵣec, the output coupling times the record-chain efficiency, both declared as intervals — and the per-event pointer-overlap deficit is computed from the device's spectroscopic parameters (2 chi / kappa = 1. 0, so the which-state information is fully exported per photon). The declared linear fit proves mathematically inapplicable at this device's measurement strength, so the extraction is executed as a documented protocol-deviation conservative bound — a declared-tolerance ceiling read against the source's own validated line-shape model, from text-stated quantities only, with a full sensitivity table — and the source's data show no excess dephasing beyond that model: the calibration returns not a point but the pre-declared gate-truncated band Lambdacal = 5e-5, 2. 1, computed at the conservative corner of the declared inputs (including the record-chain efficiency) so the dominant systematics sit inside the band; the two bookkeepings of standard budget and kernel excess are ledger-distinct while the physical carriers may overlap. The band is nonempty, so the normalization survives: lambdadec is bounded above, for the first time, by a declared-tolerance reading of laboratory line-shape data (a factor of five below Paper 41's illustrative interferometric ceiling at the most conservative record-chain reading, tightening to a factor of fifty under the architecture-level reading), lambdadec = 0 is never carried (the trade-off floor truncates from below), and the normalized kernel — the lambdadec band, the lambdadiff prediction band, the computed order coefficient, the unchanged thresholds — is handed to Paper 43. The retired authority is the Global Reducer: the demand that an adopted dynamics earn its standing only by a globally certified exact reduction of the established theories. Its methodological face falls to the construction — exact-but-sector-relative unitarity, on-average Newtonian sourcing over declared classes, imported-route compatibility at declared order, and declared-calibration normalization are how correspondence is legitimately held; its substantive face (one true global reduction) is declined, not refuted. Every result is conditional on Pₖer; the calibration datum is spent, never scored; no gate is crossed and no U3 is claimed.
Tomoyuki Uchida (Thu,) studied this question.