"Is it the case that certain optimization problems, such as MINLP or NLP (for example, problems involving a large number of equality constraints), are inherently formulated to be solved using exact methods, making it particularly challenging or inefficient to solve them using metaheuristic approaches, even for small problem instances? Or is it expected that every small optimization problem should be solved efficiently, regardless of whether exact or metaheuristic methods are employed?"