11 / C3-I: The Frontier UV Cap, the Critical-Defect Trichotomy, and One-Step Ancestry Decoupling

C3-H arrived at the obstruction that "the packet anchor is compact, but the full critical field is not." This round switches to a more structured zoom: the first frontier crossing $T_Q=\inf\{t:\exists q\ge Q,\sigma,\ a_q^\sigma(t)\ge\beta_\ast\}$ — at this moment, every shell above the frontier has not yet exceeded the threshold. The Frontier UV Cap Theorem (Theorem 8.1) gives a one-sided Besov-type upper bound, $\sup_{j\ge0,\sigma}2^{-j}\|\Delta_jP^\sigma V_Q(0)\|_\infty\le\beta_\ast$ — but citing the Seregin theorem together with critical scale invariance, it simultaneously proves $\|V_Q(0)\|_3\to\infty$ (Theorem 10.1). This produces a core tension: the finite frequency band and finite spatial core above the frontier are necessarily bounded (Theorem 13.1, Finite High-Side Core Bound), yet the global norm diverges, so the triangle inequality directly yields the Frontier Defect Trichotomy (Theorem 15.1): the divergence must come from a relative IR reservoir (D-IR, below the frontier — exactly the source side of the causal parents), UV multiscale multiplicity (D-UV, infinitely many shells can still accumulate beneath the shell cap), or spatial multiplicity/escape (D-SP) — stronger than C3-H's merely formal listing, because congestion in a finite-frequency, finite-spatial core above the frontier is now genuinely ruled out. The paper's most important positive result is the One-Generation Defect Decoupling Theorem (Theorem 21.1, conditional): under the assumptions of eventual local-source dominance plus a lower bound on phase/locality efficiency, even if the global defect diverges, the fixed-proportion nonlinear source that a child needs for its first threshold crossing can still be supplied by a finite phase-space core independent of $Q$ — the globally divergent critical background can be decoupled from the direct causal-source dynamics. But the document explicitly warns this is only a "one-generation/one-window" decoupling, not a "dynamically invariant" decoupling, and does not rule out the defect re-entering the core later on. Two explicit counterexamples support the reality of this tension: a spatial-multiplicity scalar model (§26, a superposition of unit-scale packets separated by large distances, with shell $L^\infty$ bounded but $L^3\sim N^{1/3}$ diverging with the number of copies) and a UV multiscale scalar model (§28, a family of separated critical packets, with a uniform single-shell cap yet $\|F_M\|_3\sim\beta M^{1/3}$ diverging) — both explicitly flagged as abstract counterexample ledgers, not N–S solution constructions, proving only that finite energy plus a shell cap alone is not enough to rule out this kind of multiplicity growth. The research frontier shifts to C3-J: if a distant defect repeatedly re-enters the genealogical core later on, must it pay a quantifiable cost?