Sorry for the late response - yesterday was a crazy day IRL. I think my reply here is too long, so I’ll have to break it into two comments.
You have quite a few serious misconceptions about linguistic concepts in general here, so I’m going to try to go topic by topic to address them, and hopefully by the end the idea will become clearer:
The Regularity of Sound Change
The first serious misconception is your idea that “sound change isn’t that regular”.
It was discovered in the second half of the 1800s by a German group of young researchers called the Neogrammarians (Junggrammatiker) that sound change is in fact completely regular and exceptionless in its environment. This is called the Neogrammarian Hypothesis, or just “The Regularity of Sound Change”, and it’s the foundation for all of historical linguistics (as well as being the advance that led to the discovery (by Jan Baudouin de Courtenay and the rest of the oft-ignored Kazan School in the late 1800s) that synchronic phonology is regular as well - which, by the way, is exactly why sound change is regular). Linguistic reconstruction, and even proving that languages are related to each other, is not possible without the assumption of the regularity of sound change (for reasons I can explain if you’re interested), and data from around the world has borne out the hypothesis again and again.
Note that when looking at historical data, you will undoubtedly find what appear to be exceptions to the regularity of sound change. These exceptions always occur after the sound change, and are always due to either: 1. Analogical processes or 2. Borrowing/Lexical Innovation.
Analogy
- Often, a sound change will occur (exceptionlessly, because that’s how sound change works), but sometime later the individual will imagine some sort of resemblance/relationship between one of the words sound change affected and another word or set of words, and will remake that individual word in their mental lexicon to fit the pattern they are perceiving. This is analogy, of which there are many different kinds. The remaking of femelle as female on the pattern of male, for example.
Note, however, that the initial sound change will have been perfectly regular in its environment (because, again, that’s how sound change works), and the later analogical processes have created what only appear to be exceptions to the sound change in question.
(As an aside, it’s worth noting that once analogy (or borrowing) begins to create forms that violate the environment for the original sound change, we can conclude that the phonotactic constraint that led to the sound change in question is no longer active in that language’s synchronic phonology. This is the answer to your question: “When did it stop ease the articulation?”)
Borrowing
- A language will often borrow words from other languages or from other close dialects that show different sound change outcomes (compare the native Latin ruber vs. rufus, both meaning ‘red’, but the latter clearly borrowed from a closely related Italic language that underwent dh > f in this environment instead of dh > b - again, regularity). This can also create what looks like exceptions to sound changes, especially in borrowings from dialects closely related to the dialect that underwent the change.
(This is often a better test for when “easing articulation” stops than analogy - if a language can borrow a word or alternation with that pattern, then that pattern must not be disallowed by that language’s phonotactics any longer.)
If a borrowing happens before a sound change, that borrowed word will undergo the sound change just as any other word of the language will, but if the word is borrowed during/after, it will only undergo the change if the phonotactic constraint (the synchronic realization of a sound change) is still active.
An Example
On to an example. I’m so glad you picked “listen” - it’s perfect for our purposes. This sound change is completely regular, as it turns out; it’s the change that gives us silent "t"s in listen, and soften, and fasten, and whistle, and thistle, pestle, castle, and many others. The environment is easily defined: t > 0 / 'VF_l/n (that is, “t” is deleted after a stressed vowel and a fricative, and before a syllabic (or schwa-supported, if you prefer) n or l, and every single word that fit this pattern at the time this phonotactic constraint was active underwent this change, without exception.
Now, you might very well ask, “What about often? It perfectly fits this environment, but in my dialect (maybe) it’s pronounced with the ‘t’!” What happened here is what’s called “spelling pronunciation”, which is a type of Analogy.
Once the phonotactics of the language have changed (after all of these 't’s have been removed from the language, the phonotactic constraints of the language changed, and these sequences were allowed again (there just weren’t any present for a while - an “accidental gap”)). Then, speakers a few generations later began to pronounce “often” specifically with a ‘t’ due to its spelling (likely in a misguided attempt to sound more “correct”), and we now have what appears to be an exception to a perfectly regular sound change, even though it’s not really an exception - the sound change affected ‘often’ just the same, but then another process came along and changed it afterward.
(Note that only “often” is affected by this analogical change - analogy is irregular and unpredictable in its effects, unlike sound change. This is because analogy is a lexical process, while sound change is a grammatical process.)
So, to sum up: Yes, sound change is perfectly regular in its environment, and if it looks like it isn’t, then either a) You have the wrong environment b) Analogy has affected the output of the change or c) The apparent exception is a borrowing or later creation of some sort. These are all of the possibilities - there aren’t any others.
Note that what you’ve said here:
Language change often effects some words, or a single one but not others.
is technically correct, but only because you’re conflating sound change and analogy. Sound change is regular, while analogy is not. (Check out Sturtevant’s Paradox for more about this - it’s fun, though it’s a bit orthogonal to our discussion here.)
Productivity and Regularity in Synchronic Phonology
In the same way, synchronic phonology is also regular, and describable through the same sorts of rules as diachronic phonology (though we should note that these are not describing the same object - synchronic phonological rules describe processes happening in a single human brain, while diachronic sound rules describe relationships between grammars that exist at different points in time, a meta-analysis, hence de Saussure’s famous argument about the primacy of synchrony over diachrony).
What this means, in the context of the current conversation, is that if, as you say, the “phonetic easing” process is still active in modern English, you need to be able to provide a regular, exceptionless environment that can describe it.
You’ve attempted to do this to some degree with your consonant deletion examples (even if your proposed pronunciations for strong “the” and “to” are pretty dicey), but in order to prove that the sound law that produced the a/an alternation is still a regular phonotactic constraint in Modern English, you’ll have to provide a regular synchronic sound rule that can exceptionlessly describe the phonetic environment of the constraint in question that leads to the deletion, which I don’t think you’ll be able to do.
Note that your proposed rule must not be specific to individual lexical items or refer to morphological or syntactic boundaries. This is because:
Structure is not Visible to Phonology - the Modularity of the Grammar
It’s traditionally assumed by most generative linguists that the grammar is largely modular - that is, each phase of the generation of an utterance is separate, and proceeds one at a time with little overlap between the modules. So, syntax first builds the structure of the clause, and then morphology (which does not have access to the syntactic structure (though see Distributed Morphology for modern attempts to unite syntax and morphology)) builds words to fit into the structural positions that syntax built, and then phonology (which similarly cannot see either syntactic or morphological structure) determines the sounds that are sent off to be pronounced by the articulators. (Note that the actual relationships are a bit more complex - see Kiparsky’s 1982 book on Cyclic/Lexical Phonology for a famous example that’s pretty accessible, but the generalization holds well enough for the data we’re dealing with here.)
What this means is that synchronic (and diachronic, for that matter) sound rules only ever apply in phonological environments, that is, to strings of phones and suprasegmental features like tone, stress, etc. (which does include prosody), and not to individual words.
So, in order for the “ease of pronunciation” constraint you’re referring to here to still be active in Modern English, it must be describable as a phonological rule that applies exceptionlessly in a specific phonological environment, regardless of the words or structures that are actually present.
This is why I don’t think you’ll be able to show that the a/an alternation is still a regular, productive alternation in Modern English. The a/an alternation is not predictable - there is no general rule in English phonology of which its behavior is a subset. A child acquiring English just has to learn that for this specific morpheme, there’s an “n” before vowels and no “n” before consonants, and, crucially, no generally describable phonological sequence in the language works this way.
We can test this with the analogy and borrowing tests above. First, through the analogy test, “my/mine” no longer behaves this way, because its behavior has been altered through a combination of analogy and grammaticalization - the sound law clearly no longer holds in its environment, so the phonotactic constraint that produced it is no longer active in the language. Second, and this is admittedly a hypothetical, I don’t believe that any new monosyllabic word borrowed into English ending in -an (or -uw or -ij, for that matter) would show the same alternation in any environment, which would again indicate that the phonotactic constraint is no longer active.
All of this is because the regular sound change that originally produced this alternation is really just as fossilized as the medial f/v alternation: neither alternation can be successfully described using exceptionless synchronic sound rules, and must therefore be stored in the lexicon (“fossilized”) and learned as exceptions by new acquirers.
(Note: Both of these alternations are morphologically/lexically-conditioned allomorphy, if you’re interested.)
I hope this makes sense. Sorry if this was way too much info - it felt nostalgic, like being back in front of my third- and fourth-year undergrad students again, and I got a bit carried away. Also, I like your username. :)
Let me know if anything here is unclear or if you have further questions.