No one knows, nor could they ever know.
Given enough of the right samples and their context, you can make strong informed guesses about pronunciation from how words are used in structured samples like poetry/plays that embed rhythm, rhyme, etc in established ways; from patterns in the way words shift over time; etc
There are quite a lot of signals you can feel confident drawing from, but of course you can't get to 100% accuracy even if you had perfect insight or even heard the word spoken because pronunciation does still vary by individual, family, community, context, etc
I don't know what's right here myself or how established the consensus is around what might be right, but the discussion you're responding to isn't frivolous or unreasonable on its face.
The consensus might be bang on. Or it might be wildly wrong. Or something else. Who can say?
My point is a pretty straightforward one. To say you have knowledge of a thing, you need to be able to personally verify whatever-it-is. I can personally verify I'm sitting in a chair. I can never verify that music sounded like that 2000 years ago, cos I was not there. I can read a book that describes the sound, but that still doesn't mean I can verify it!
"We" cannot know a thing - there may be a consensus opinion (the wisdom of crowds and all that) - it might be a good guess - but this is not knowledge. And can never be.
This doesn't take away from the fun endeavour though. If you want to recreate sounds from the descriptions given in text, go ahead - I'll listen to it.
By your standard, most of us also don't "know" that the Civil War happened or that there is a vacuum in space. It's deductions, inductions, personal attestations, and consensuses all the way down. Yet, to navigate a modern world that presents itself as scientific and evidence-based, most of us adopt beliefs about these things -- sometimes with much stubbornness and sometimes very lightly -- and call these things "knowledge" too.
You can be a stickler about your use of the term, but the more you are, the more distanced you find yourself from common usage, the more needlessly contentious and pedantic discussions become, and the more you're left to come up with your own language to describe all the usages you've chosen to reject.
On the other hand, my assessments and understanding is my own. My brain is not "outsourced" to consensus or experts. I'm fine with that!
I'm interested to see if you have a suggestion that would draw this distinction out better. The idea that something can be 'known' at large by a group, and the idea that the individual 'knows' because he has personally verified something. Are both these really "knowledge"? Can knowledge be a team sport? Is knowledge what the person with the biggest megaphone, or the most supporters say it is?
I'm also interested to hear a better term than 'know', for those things that the individual has verified vs what you would call (group) "knowledge" - but I personally think my 'hard' definition is the proper one. One knows, or one believes (or theorises/hypothesises/etc). It might (or might not) be fair/useful to hold consensus opinion - but, as far as the individual is concerned this type of information ought to be characterised as 'belief'.
I'm sure you can tell, but I think knowledge (all thinking and emotions actually) occur in the individual only. A generic "we" cannot know anything - knowledge is enlivened in the individual mind only.
I think this is all self-evident, but I'm always happy to hear alternative views, better definitions of terms, etc. I invite you to explain further and correct me!
There really are some things that cannot be known. It is valuable and useful to distinguish between belief/hypothesis and knowledge.
The worst thing, imo, is to believe that one knows when one is a state of belief - this is the curiosity killer.
Neither can a person who has never left the US who realistically would not visit France verify whether they are speaking French as is spoken in Marseilles. Yet the effort to approximate it is not futile, and having practiced with a native French speaker from Marseilles to the point of fluency is usually proof enough for many to say that they know French as is spoken in Marseilles, even though it seems to me that you would not regard this as knowledge.
How do you propose one verifies how music sounded 2000 years ago?