A new arXiv paper introduces a post-solve audit layer for Mixed-Integer Linear Programming engines. It identifies how small perturbations in costs or demands trigger discontinuous shifts in optimal plans. This framework provides solver-backed evidence of solution reliability. Practitioners can now quantify the trust gap between nominal optimality and real-world deployment.