Discuss Findings

mean_crowding_induced_reentrant_gain -0.00022 +/- 0.00012 (d=-1.84861, sign-consistent=2.3%); mean_interior_best_fraction 0.35187 +/- 0.00577 (d=61.03276, sign-consistent=100.0%)

SOFT REJECTION. The reentrant gain is small but stably negative (d=-1.85, 97.7% negative sign across 2647 seeds), meaning the crowding-trap memory release window produces a mild anti-reentrant effect — opposite to the hypothesis. Meanwhile, 35.2% of seeds find an interior optimum (d=61.0, 100% consistent), confirming that the parameter landscape has nontrivial structure. However, this structure generates a consistently negative reentrant gain rather than the predicted positive one. The interior geometry exists but the reentrant mechanism runs backward.