To explain and illustrate this:
Given a user of Paa.nini’s grammar who feels the need to account for the utterance of a well-educated (;si.s.ta) speaker of the bhaaSaa (a function of the grammar, not free from a certain amount of circularity, but generally accepted by later Paa.niniiyas), what knowledge should he minimally have in order to accomplish such a task? Suppose a well-educated speaker wanted to express that Gautama’s being and becoming have been very intensive (or that he “became” repeatedly, i.e., now and in previous lives). He could have said:
gautamo bobhavaaM cakaara
Did he stammer or speak correct Sanskrit?
It is well known that any user of Paa.nini’s grammar (1) should already be fully conversant with Sanskrit or what Patañjali calls bhāṣā. In addition, (2) he should be conversant with techniques and terms of grammar presupposed but not explained in Pāṇini’s grammar (kaaraka etc.). Not unimportantly, (3) he cannot proceed unless he already masters a number of lists (gaNas) that come with the grammar, including the one often mentioned separatly with ca. 2000 items: the list of dhaatus or roots.
Our user of Paa.nini’s grammar may not feel the need to derive the proper name Gautama. So let us turn straight away to bobhavaaM cakaara.
Before the grammar-user can start his endeavour to account for this form by derivation, he requires a skill which is not at all self-evident but to which his previous formation under (1), (2) and (3), must have prepared him quite well: even if bobhavaaM cakaara would not be part of his active vocabulary he should be able to propose that, formally and semantically, BHUU might be the underlying root. Once BHUU is selected the grammar will ask him further semantic and formal questions (see Paninian grammar through its Examples, vol. 2 p 607). He will then be able to re-constitute the form, and at once confirm its correctness as well as the ;si.s.ta-status of the speaker, and, in the process, become aware of the semantic niceties involved in this form apart from a simple choice such as the number to be selected (singular).
Now the SAME user of Paa.nini’s grammar with the SAME cognitive and linguistic skills (who, if he is contemporaneous with Paa.nini, will ALSO be familiar with the other end in his diglossic range, Middle Indic), will surely also be able to posit BHUU underlying in bhoti in Aśoka’s dharmacaraṇaṁ pi ca na bhoti aśilasa. This grammar user will then find that, answering some simple semantic questions, he ends up with a slightly different form, bhavati. Going through his grammatical rules and lists for each term he will sooner or later end up with a SANSKRITIZED form of Aśoka’s statement: dharmacaraṇam api ca na bhavati aśīlasya.
And WE end up with discerning a new, so far unacknowledged, function of Paa.nini’s grammar: a function that is adjacent to “accounting for utterances of well-educated speakers”, but (instead of being circular) useful and productive in the diglossic situation of Paa.nini’s time and environment. Were these ancient grammarians perhaps not as dissociated from their society as we now tend to think?
Could this new function explain or clarify the closeness of Paali to Paa.nini’s grammar (observed several times, e g O. von Hinueber in Das aeltere Mittelindisch... p. 123)?