Paper 108B extends the reduced-support QCD-like diagnostic developed in Paper 108 from a single L = 16 discovery basin into a scale-locality robustness test through L = 32. The paper does not claim to solve full QCD, replace lattice-QCD computation, derive a continuum confinement law, or compute hadron spectra. Instead, it asks a narrower mechanism question: does the TDR-anchored graph-corridor mechanism found in Paper 108 survive increasing lattice scale, and does the correct measurement location change as the transported signal moves outward? The tested mechanism combines a calibrated two-defect ring, or TDR, with local VROS support and a delayed 252 to 362 graph-corridor capacity-lag channel. The TDR supplies the compact anti-periodic boundary condition. The VROS scaffold preserves local boundary memory. The graph corridor carries the outward response. The hard diagnostic remains balanced two-step square-Wilson growth. A row is counted as a breakthrough only when both Wilson-growth steps are positive: from the 1 by 1 loop to the 2 by 2 loop, and from the 2 by 2 loop to the 3 by 3 loop. The scale sequence shows that the Paper 108 result is not isolated. In the v44 scale-calibration run, the target graph-corridor mode produced breakthrough rows at L = 12, L = 16, L = 20, and L = 24, with strongest early survival at L = 16 and L = 20. Under the original TDR-near measurement ruler, L = 24 mostly faded. The v45 recovery test showed that this was not simple mechanism failure: at L = 24, graph-corridor-near measurement recovered 13 target-mode breakthrough rows out of 45, while TDR-near measurement produced only 1 out of 45. The larger-scale tests strengthened this interpretation. A v45B L = 32 scout recovered the signal under graph-corridor-near measurement. The v46 six-seed L = 32 confirmation then produced 25 target-mode breakthrough rows out of 96, with at least one breakthrough in every seed. Full-lattice control produced no v46 breakthroughs, while random equal-support control produced one. These results support a scale-local transport interpretation: at smaller scale, TDR-near measurement can still detect the coupled boundary and handoff, but at larger scale the Wilson-growth signal is best recovered when measurement follows the graph corridor. The conclusion is bounded but stronger than Paper 108. The TDR-anchored 252 to 362 graph-corridor mechanism repeatedly produces balanced two-step Wilson growth across scale when measured near the transported corridor. The result is not full QCD, but it is a reproducible and falsifiable reduced-support boundary-corridor basin. The next decisive tests are shell-ladder versus off-ladder controls, expanded L = 32 seed and replica confirmation, TDR-family robustness, and a self-organizing admissibility-based corridor solver.