complications: hard to compute equations, if we use integral by paths approach, we are getting often infinities.
(1,2) I think it is always valid, if the correction is much lower than a value from the previous order of perturbation.
(3) I think it should be feasible, just we need to understand which is appropriate model. For example, for high energies, we can’t use quantum mechanics at all as I know.