Most real world problems have constraints that render global optimums moot. Usually opportunity cost, but also risk, experiential components, satisficing ...
In fact, any interview problem with a global optimum is worthless for evaluating candidate's ability to handle ambiguity or demonstrate capacity for creativity, second order thinking, communication ...
Expecting candidates to lead with absolutism like in your post os a good recipe for finding Brilliant Jerks.