Is there any information on the status of learned pronunciations from the late imperial period up to 1000 CE? I am wondering because the Classical Latin reconstruction seems to make clear that by the imperial period, final /m/ was regularly lost and replaced with nasalization on the previous vowel, and then of course there are the near-close short vowel allophones [ɪ] and [ʊ] for /i/ and /u/. But while in Vulgar Latin the graphic omission of 'm' indicates the lost of nasalization, and the 'e/i' 'o/u' confusion indicates the vowel mergers, later medieval Ecclesiastical Latin seems to follow a learned spelling pronunciation according to the local Romance dialect, restoring the consonantal status of final /m/ and pronouncing former short vowels the same as their former long counterparts, so [ɪ] and [iː] > [i] and [ʊ] and [uː] > [u] instead of [ɪ] > [e], [iː] > [i] and [ʊ] > [o] and [uː] > [u] in VL.
I presume that learned speakers would have followed the general patterns of most other VL./Proto-Romance consonant developments, so pre-palatal /k/ > /tʃ/ or /ts/, /g/ > /dʒ/, lost of /h/, etc. But my problem is, when would the restoration of spelling pronunciations for vowels that we see in medieval Latin––like final /m/, near-close vowel elimination, etc.––have taken place? For example, while a Classical Latin speaker would have pronounced ad dominum as [ad ˈdɔmɪnũː] would Jerome or Boethius have already said [ad ˈdɔminum], as in later ecclesiastical Latin, while their contemporary Vulgar Latin/Proto-Romance speakers would have probably said something like [a ˈdɔmnu] or [a ̍dɔmno]?