Nice article. I think Lakatos’ point in refuting Popper’s naive falsificationism is related to the point that observations are always theory-laden.
Quine also said something like the theory as a whole is composed of many propositions, (including background assumptions), and it is logically impossible to isolate any of the propositions and subject it to empirical testing (see Duhem-Quine thesis).
Anyway, all of these concerns whether observations can verify/falsify a theory 100%, i.e. in an all-or-nothing fashion. A Bayesian approach seems like a good idea to address our change in belief in a theory or proposition, given new evidences.