Cube Domination: Coordination-Frame Collapse, Modular Decomposition, and the Re-emergence of a Dominant Frame

Multi-agent systems sharing a coordination frame can face increasing task interference and a trade-off between modular separation and coordination overhead. This paper connects three explicit submodels: decomposition cost, selection of a new frame under survival and mean-coupling constraints, and distorted observation. Frame failure is an initial condition; autonomous collapse and persistent post-selection dominance are not derived. Under consistent pair-counting costs, the paper derives balanced-allocation criteria, finite-size corrections and a constructive sufficient condition for advantageous separation on nonnegative conflict structures. Relaxed and exact integer optimizers can differ despite vanishing normalized value error. A common quality response cancels from selection at fixed elapsed time and fixed active candidates; with heterogeneous response, surviving first-hitting completion is characterized exactly in the two-candidate baseline. Radial dilation connects that condition to geometry; known initial distances give an exact mean-coupling gate, and response intervals admit a uniform survival test. Co-dilated access costs can be included without requiring globally monotone support. An assumed affine transmission law connects decomposition gain to reduced dose and preserved survival feasibility; the law remains unverified. Exact class invariance is separated from asymptotic and finite-size claims. Bounded noise, suppression and calibration give pointwise decision sets and a joint consistency test when realized dose increments are independently known; conditional alarm ordering remains distinct from practical warning performance. There are no numerical experiments. Predictions specify independent calibration, prospective denominators, competing mechanisms and measurable rejection conditions. The formal results hold under their stated models; external applicability, autonomous attainment and recursive-layer closure remain separate questions. Note on Version 2.0: this version revises the registered v1.1 (title then: "Re-emergence of Dominant Coordination Frames Following Global Collapse"). The numerical toy runs distributed with v1.1 are no longer part of the submission and no numerical experiment is claimed; the argument is restated as connected analytic submodels with counterexample and candidate ledgers, and one structural sensitivity claim is explicitly withdrawn. Files: the revised manuscript (v2.0) and a support archive containing the manifest and SHA-256 checksums. The v1.1 record and its scripts remain available under the same concept DOI.