IRMT XVII - Compact Determinant Closure and the Finite Charged-Lepton Triplet: Winding Records, Naturality Forcing, and Completed-Record Protection of the Determinant-Line Metric

Paper XVI of the Information Relational Manifestation Theory program asked whether the response-connection law required by the charged-lepton construction follows from the microscopic substrate dynamics. It concluded that it does not. The law relating the stabilized packet response to the determinant-line current was retained as the minimal constitutive endpoint of the effective theory, and the construction of Paper XV was reclassified as a highly constrained conditional effective model rather than a microscopic theorem. Paper XVI left three logically distinct paths: accept the constitutive law, introduce new microscopic protection structure, or derive a deeper packet action. This paper reports that the second and third paths were both taken, and that the constitutive endpoint is now closed within a declared finite and countable domain. The closure does not repair the local route audited in Paper XVI. It replaces that route entirely. The mechanism turns Paper XVI's own obstruction into the instrument of closure. A globally smooth periodic scalar on a compact phase circle cannot carry a nonzero constant derivative, so the exact compact completion must instead carry an integer winding record. Once that record is admitted, three independent structural requirements, namely cross-dimensional unitary equivariance, orthogonal direct-sum compatibility, and elementary-record normalization, collapse the space of admissible response-to-current transformations from eighty-one real dimensions to two, then to one, then to none. The surviving transformation is the identity. The required law is therefore not chosen but forced: there is nothing left to choose. Three of the eight objects Paper XVI listed as underived, the identity intertwiner, the unit lift coefficient and the compact phase normalization, follow from this classification rather than from assumption. The second obstruction falls to a different argument. When the protected projectors are treated as completed retained records, admissible future events intertwine each projector individually, and the second exterior power then preserves the determinant-line subspace and the mixed-wedge sector separately. The cross block vanishes identically, so the Schur correction does not exist on the admissible tangent space and the determinant-line metric is protected exactly rather than suppressed numerically. Paper XVI's quantitative decoupling requirement is therefore not met; it ceases to apply. We recompute that threshold independently and reproduce it to within a part in a thousand, together with Paper XVI's mixed-wedge audit values to ten decimal places, in order to establish that the two papers describe the same object under different admissibility conditions. Conditional on the closure, the charged-lepton chain becomes a finite numerical prediction. With the Fermi constant as the sole empirical input and no charged-lepton mass used anywhere in the derivation, all three charged-lepton masses are reproduced to better than three parts in ten thousand. The three deviations share a sign and nearly a magnitude: a single common rescaling of the packet scale by two and a half parts in ten thousand brings every residual below four parts in a hundred thousand, and the predicted square-root mass ratios match measurement to five parts in a million and to three and a half parts in a hundred thousand. We report this decomposition as a diagnostic that localises any required correction in the scale bridge rather than in the Koide geometry. It is not a second prediction, and the common factor is a one-parameter fit made after the fact. The result is closure relative to a disclosed premise bundle, not unconditional derivation. Five premises do the work: the winding-record completion of the compact phase; cross-dimensional naturality of the response transformation; elementary-record normalization; the s