This model is explicitly requiring that you deal only with physical processes, so your convex function solver would require time to get from the starting state to the end state. If it is happening non-iteratively then it would cease to be an optimizing system after it has completed the function, since there is no longer a target configuration.