A minimum viable pool of 200 diploid founders, a frequency cap retaining 0.0707 of the uncapped benefit, and replicate-mean heterozygosity retention of 0.9686 through a founder funnel matching its closed-form prediction under Amendment A-S4: three pr

Three pre-registered simulation studies, each run on the same five-locus panel seeded from gnomAD allele frequencies in 30 replicates under criteria frozen before any simulation ran, measure three operating parameters of a synthetic genetic pool: the smallest governable cohort, the price of a diversity constraint, and the effect of a founder bottleneck. The minimum viable pool is N = 200 diploid founders, the smallest tested size meeting the amended criteria, determined under a corrected fidelity statistic after the frozen tier was voided by a failed sanity gate. A frequency cap of 0.02 per locus per generation, implemented by a projected-cohort rule that can test the completed cohort it constrains, holds in 179 of 180 replicate-generations and retains 0.0707 of the uncapped benefit, so about 93 percent of the benefit is surrendered; the frozen greedy rule retains 0.0287 and is a non-instance of the cap it was meant to price. Heterozygosity retention through a founder funnel is 0.9686 against the closed-form prediction of 0.9654 under a mean-based criterion declared by amendment; the frozen median-based verdict is a fail and stands unrevised. Each correction is declared in a dated amendment and reported as a separate labeled tier, and no frozen verdict is revised by any later tier. The manuscript introduces no new data and no new computation; every number traces to the deposited results files of the three source studies.