In a seventeenth century world the Newtonian model did mostly very well to describe how gravity works in the universe and did well with most empirical evidence of that time. Of course now we know that gravity is not that Newtonian but described more correctly by General Relativity. My question is does there exist an ad hoc hypothesis that we could add to the Newtonian model to arrive at the Einsteinian model and more generally can we always find an ad hoc hypothesis to result in a more sound generalization ? Because if we can I think this will violate the heuristic of parsimony that the simpler theory is more likely to be correct since this likelihood decreases as we have more chances of finding an ad hoc hypothesis that will result in a more general theory. Does this mean a particular over-fitted extrapolation (which would contain that particular ad hoc hypothesis) out of many such over-fitted model will stand a chance of predicting with higher accuracy thus disproving the bias-variance trade-off ?
My question is does there exist an ad hoc hypothesis that we could add to the Newtonian model to arrive at the Einsteinian model
No. General Relativity is not Newtonian mechanics plus a bit. Newtonian mechanics is a special case of General Relativity.
Now, if you question was:
does there exist an ad hoc hypothesis that we could add to the Newtonian model that produces the same results as the Einsteinian model
Then the answer is: possibly. Or more precisely, one may be able to find an adhoc hypothesis that, for most practical purposes, makes the same corrections as General Relativity. I don't believe this approach was ever taken with General Relativity (Einstein was attempting to reconcile Special Relativity and Newtonian Gravity, not explain anomalous experiments) but is fairly common in physics and is known as phenomenology. It is considered to be somewhat of a hack, though, and not a good substitute for a more general theory.
can we always find an ad hoc hypothesis to result in a more sound generalization ?
Historically, this has rarely been the case. It has been much more prevalent that a generalization has been found for which the previous theories were special cases.
this will violate the heuristic of parsimony that the simpler theory is more likely to be correct
But that's ok. The bit that's often left out with Occam's Razor is the "all other things being equal" bit. It's only useful as a tie breaker if the hypotheses you're comparing have equivalent explanatory power. As Einstein (may have) said, a theory should be as simple as possible but not simpler.