Cosmological Boundary Routing

All three canonical direct ladder encodings — classical Friedmann contraction (A), post-bounce LQC expansion (B), and Planck-anchored ΛCDM (C) — return FULL_PERCOLATION with final giant ratio = 1.000 and isolated = 0. The result holds even for Dataset B, whose expansion ladder requires a connectivity threshold approximately 7.5 times larger than A or C to assemble. This is not a failure of B: it indicates that the selected post-bounce LQC expansion encoding carries a structurally broader gap hierarchy than the matched classical encodings. All three canonical direct encodings preserve A_κ = 1.000 across the tested STRUC-I κ range. The full Dataset B route also remains overwhelmingly admissible under α deformation (98.83% admissibility persistence), with all localized instability confined to symmetric near-bounce shells and none at the exact turning surface.

Dataset B's sorted expansion ladder requires κ_connect = 0.1000 to form a single connected structure, versus 0.013335 for both A and C. This 7.5-fold elevation is not an admissibility violation — it is a structural distinction. The selected post-bounce LQC expansion encoding carries a broader gap hierarchy than the matched classical encodings. This is consistent with, but does not by itself causally isolate, structural inheritance from the preceding bounce transition; the α analysis, not the direct ladder, localizes behavior around the quantum-transition shell. At the smallest tested vulnerability scale (κ = 0.01), A and C are already dominated by a giant component containing approximately 97.675% of the ladder (77 components each), while B's expansion branch is substantially more fragmented (GR = 0.53277, 22 components). Full connectivity for A and C is achieved at κ = 0.013335; Dataset B requires κ = 0.10000.

The admissibility metric A_κ = 1.000 holds at every κ point for all three canonical direct encodings. No direct ladder enters a non-realizable structural state. However, Dataset B's mean structural pressure (ρ̄ = 0.306) exceeds A and C (ρ̄ ≈ 0.042) by a factor of approximately 7.4. Its maximum reaches 0.636 compared to 0.290 for A and C. The STRUC-I classification of Weak Persistence (versus Stable Structure for A and C) captures this: B remains admissible but lies closer to the structural boundary, operating under substantially higher internal deformation pressure. This elevated pressure is measured in the post-bounce LQC expansion ladder and reflects its broader sorted gap geometry. Note that B's direct STRUC-I uses only the 2000-row post-bounce expansion ladder; the near-bounce spatial localization belongs to the full-path α analysis.

The 982 unstable rows in Dataset B are distributed symmetrically: 491 in the contraction near-bounce region and 491 in the expansion near-bounce region. Zero unstable rows occur at the exact bounce-crossing intervals. The outer contraction and outer expansion regions are entirely stable. This pattern identifies the quantum-transition shell (approximately 0.03 ≤ ρ/ρ_c ≤ 0.82) as a narrow deformation-sensitive region surrounding, but not including, the turning surface itself. The collapse onset radius of 0.05 is a strict first-local-instability coordinate: localized near-bounce sensitivity begins at a five-percent deformation, while 98.83% of the full grid remains admissible. The LQC bounce is not structurally forbidden — it is admissible at the exact turning point and broadly stable away from the transition region.

The combined experiment produces one route topology result readable in a single dimensionless coordinate. Dataset A (classical Friedmann contraction) approaches a terminal boundary: scale factor decreases, density increases, provisional margin decreases, and the represented classical route terminates at coordinate zero, with no continuation included in Dataset A. Dataset C (Planck ΛCDM expansion) recedes from the corresponding early boundary regime: all signs are reversed relative to A. Dataset B (effective LQC) exhibits neither pattern — it approaches a finite turning locus at coordinate zero, preserves path continuity through the route change (zero unstable rows at the exact bounce crossing), and continues onto the opposite branch. The three route classes — terminal approach, finite route reversal, boundary recession — are computationally distinguishable within this controlled pilot corpus. Magnitude-only analysis cannot produce this distinction; orientation-sensitive route comparison is required.

Stellar Boundary Dynamics identified branching routing: a catastrophic event generates internally admissible but mutually non-equivalent downstream regimes. All three stellar phases returned FULL_PERCOLATION, but the ABC bridge revealed A→B contact with C branching — demonstrating that internal admissibility is not cross-regime equivalence. The cosmological experiment identifies a complementary morphology: turning-locus routing, in which a continuous trajectory approaches a finite critical locus, reverses orientation, and remains admissible through the route change. The identity-preserving path is not broken; it is re-routed. Together, the two corpora support a broader Boundary Routing framework while retaining distinct topological signatures — branching downstream in the stellar case, continuous reversal in the cosmological case. The route topology changes while direct structural admissibility is preserved across all canonical encodings and full-path α admissibility remains overwhelmingly persistent; localized sensitivity is confined to the near-bounce shell rather than the exact turning locus.