Sure, it isn’t the definitive proof because there can’t be any definitive proof either for or against it. It is a probabilistic proof that refers to a word that has been used in hundreds if not thousands of suttas and is part of the core of the Buddhist tradition. And it is a strong probabilistic argument, because the Sinhalese were very consequent with the words like dhamma (dam , dahama), sangha (sanga), bhuddha (budu), bhikkhu (bik) and countless others. So far, as far as I know, most if not all transmission errors that are agreed upon are not systematic in the sense that they show up in no more than one, two, okay, maybe ten suttas (even though I don’t know of any such examples). In case of anatta, any contexts where ‘attha’ supposedly stood for it should have been systematically purged by the reciters and scribes in so many texts, that even without any additional arguments it seems pretty improbable. I mean, this is a strong claim and it requires strong evidence, so any plausible probablistic argument against it is pretty much like the last straw, and my argument is not the only one I presented: for the sake of brevity and because I didn’t have that much time I didn’t even present arguments from the Canon in other Prakrits like Gandhariand Sanskrit (see p. 44, pp. 204-211 in this paper by Bryan Levman ans pretty much the entire part about the possible earliest language of the Buddhist tradition).
Anatta is so central and so grammatically and phonologically lucid both synchronically and diachronically that assuming an error in its transmission is similar to saying that Jesus Christ was Cheesus Christ all along, all worship the cheese! (Note: it is a joke, of course)