nobody talks about how weird it is that a common way for phonemic distinctions in languages to develop is when there's a single phoneme that has different allophones in different context and some aspect of the conditioning environment for an allophone disappears but they keep saying the allophone in words that used to have the right environment and therefore they're now in distinction
like why do we keep using the old (former-)allophone instead of changing to what would normally be expected in that environment.
Lady, EVOLVED 
@alyssa generally, because auditory communication is lossy, and the more information you can convey to help disambiguate the better
the more sounds you have, the fewer words have each sound, and the easier it is to select which word you’re hearing
this comes at a cost, which is why mergers also happen, but oftentimes the distinction is useful or even necessary, so it sticks around
(it would be a big problem if “like” and “lick” became homophones in English just because we no longer say the final E, as an extreme example)
@alyssa i think this is a difficult question to answer because frequently we only know about these things because someone considered it distinctive enough to remark upon, which is usually not the case for ordinary allophones
much of our knowledge about historical phonology comes from things like poetry, but oftentimes poetic rhyming conflates various allophones (and if it doesn’t, that’s an argument in favour of the sounds being distinctive)
so “people used to say this a few different ways, but that variety vanished without consequence once the conditioning environment shifted” is something there is likely to be zero historical record of (except in the past maybe hundred years where people have started getting really rigorous about recording dialect and phonological systems exactly)
@alyssa (so, no i don’t know of any cases, but moreover i don’t know HOW we would know)
@Lady the indo-aryan languages seem like the main spot where it might plausibly be observable (due to combination of being one of the most studied families in historical linguistics and of having a very ancient literature on phonetics)
but yeah more modern synchronic observation seems more likely to turn up cases if it exists, and certainly we know of some ongoing cases of loss of conditioning environment resulting in phonemic split of former allophones from synchronic observation.
@Lady ah that corollary's an interesting point. see, this is why books about historical linguistics should talk about this issue :P
do you know if cases are known where a conditioning environment for allophonic variation was lost and the former allophonic variation was also simply lost rather than splitting phonologically?