Floating point calculations are intrinsically subject to error. In the context of integer programming, these errors may lead to incorrect decisions in the branch-and-bound process, ultimately leading to incorrect optimal solutions. The typical approach is to use heuristic rules to mitigate those issues. This paper proposes a computationally efficient way to guarantee correctness of integer programming bounding decisions within the context of branch-and-price formulations of the vehicle routing problem.