Whence cometh rhotic vowels, if not rhotic coda?

[M Mira]

As you may have guessed from the title, this is mainly concerning Mandarin, which gained a rhotic vowel [ɚ] without having any previous liquid coda in the language.

The usual theories claim that they came from alveolo-palatal nasal onset /ȵ/ merging with the nucleus vowel, but this sound change seems to be sporadic, with others ending up as initial [ʐ/ɻ] or [n] with the nucleus vowels retained instead. What are the mechanisms behind these changes?

[Dewrad]

As you may have guessed from the title, this is mainly concerning Mandarin, which gained a rhotic vowel [ɚ] without having any previous liquid coda in the language.

The usual theories claim that they came from alveolo-palatal nasal onset /ȵ/ merging with the nucleus vowel, but this sound change seems to be sporadic, with others ending up as initial [ʐ/ɻ] or [n] with the nucleus vowels retained instead. What are the mechanisms behind these changes?

"Feature spreading".

[M Mira]

As you may have guessed from the title, this is mainly concerning Mandarin, which gained a rhotic vowel [ɚ] without having any previous liquid coda in the language.

The usual theories claim that they came from alveolo-palatal nasal onset /ȵ/ merging with the nucleus vowel, but this sound change seems to be sporadic, with others ending up as initial [ʐ/ɻ] or [n] with the nucleus vowels retained instead. What are the mechanisms behind these changes?

"Feature spreading".

So some of the nasals became rhotics first, then these rhotics, under certain conditions, assimilates the following vowels into rhotic vowels?

[Nortaneous]

Your question is: why did MC *ȵi sometimes become Mandarin ri and sometimes become Mandarin er?

I don't know. As for *ȵ > ʐ, the intermediate step that's usually reconstructed is *ȵʑ. Why did the palatal nasal spontaneously eject a fricative? Well, languages in that area tend not to make strong distinctions between fricatives and approximants, so it's basically *ȵ > *ȵj. Then palatals became retroflexes.

(Why is *ȵʑ reconstructed? Why couldn't it have gone straight from *ȵ to *ʑ? ɲ > j doesn't seem like that odd a change. j > ɹ supposedly happens in some AAVE.)