When the optimization is completed, the optimizer returns either the best parameter set (design) or the set of the best designs. Numerically, these designs are the best, however, they have to be checked as to feasibility, or even if they are physically reasonable. Actually, the constraints and barrier function should have guided the optimization algorithm towards finding more reasonable designs.
Numerically reasonable results can often be obtained very quickly. But due to physically constraints, some of these results have to be rejected. In recent times, even physically reasonable designs have to be rejected because they do not fit into the economical and financial constraints. In that cases the constraint functions were not designed appropriately and a redesign of the complete optimization setup is necessary.