Quasi-periodic prefactor at golden slope: period-2 structure, Zeckendorf-phase clustering, and separation from two controls

Erratum: the file attached to this version is the DRAFT build of the manuscript, attached in error. The deposit build is attached to version 1.1 and is reachable via the concept DOI 10.5281/zenodo.22739008. The mathematics is identical; the builds differ in draft markers only. This record reports the prefactor at the GOLDEN slope. It is a companion computation to the A076227 software deposit (Zenodo 22121363), on a different slope and a different sequence; it is not a new version of that record and retracts nothing in it. The same exact-integer dynamic programme is run at the golden slope 1/phi, where the prefactor exhibits an exact period-2 limit cycle along Fibonacci indices (last within-parity difference 0.001588, against 0.062063 for the gamma control through the identical code path), clustering of the detrended profile by Zeckendorf phase (eps_5/eps_1 = 0.198649 against a sealed threshold of 0.5), and separation from two controls. Relation to Zenodo 22121363. That record's subject is the sequence A076227 at the single slope gamma = 1/log2(3); its title, its README heading and its keywords name that slope and no other. The barrier used here is a different barrier and generates a different sequence: B(39) = 3611535862 there, 4574825460 here. The machinery is reused; the object is not the same object. This record therefore cites that one rather than versioning it. Cited, not derived: the barrier word's recursion s_(n+1) = s_n s_(n-1) is classical in combinatorics on words (Lothaire; de Luca). A parity-dependent "seam correction" that an intermediate draft derived is the translation cost of an index convention, not a phenomenon: on the classical characteristic word the identity holds exactly, with no correction, for all 14 of 14 tested scales. That was settled by construction, not by literature search. A further claim from that draft — that the prefactor's period-2 cycle and the seam defect were one fact — is WITHDRAWN. NOT CLAIMED: a solution to the irrational-slope problem; any result about A076227 itself. OPEN, named in the body: the observed decay rate 0.6676 against the predicted 1/phi = 0.6180, and a sup plateau of 0.093777 quoted in the working record which does not reproduce (the value derivable from the artifact is 0.094343; the discrepancy is recorded, not reconciled, and nothing rests on it). The prior-art scope is stated at what the gates actually covered, including that one sweep's arXiv arm screened zero documents while reporting itself operational.