Let's say there's a language where where is no phonemic voicing distinction in stops, and little or no phonemic distinction between stops and corresponding fricatives.

Question #1: How likely is it that the phonemes in question have *only* voiceless stop allophones? From what I've read, it appears that languages like that tend to have at least voiced stop allophones for those sounds; but I know I haven't read anything about allophones in Polynesian languages, which I think would be useful data.

I also wonder how long the lack of voiced and/or fricative allophones would last. This *may* in fact be the same question!

Then let's say those stop phonemes develop voiced and fricative allophones:

/p/ [p b p\ B]
/t/ [t d T D 4]
/k/ [k g x G]

#2: Now, how likely is it for the language to go back to a state where all (or most) of those non-voiceless stop allophones have dropped out of use?

There's a philosophical issue here -- how would anyone know, later on, that the language had ever had allophones for the stops? Well, in the system I'm thinking of, there would be mergers (perhaps analyzable as hypercorrection) in some places between pre-existing* non-stop phonemes and stop ones. E.g. a pre-existing /w/ might merge in some words with /p/ through the similarity to the latter's allophone [B], but in later years that /p/ would only be pronounceable as [p].