One category of problem might have the feature that it’s not too expensive to create a set of acyclic graphs that make an epsilon cover of the feasible region of the problem while evaluating proposed solution points is more expensive (for computation or for real). There is no connection between local minima because the feasible region itself is not connected. Do we expect any kind of similar phenomena in the unconnected regions? That’s additional theory. My point is that the connection between studied problem forms & actual operations research frontiers might be very loose or almost non-existant.