You bring up a good point about the reliability of answers and author reputation.
Sometimes I see answers that sound authoritative but offer no supporting evidence, or, as you are deftly not putting into these words, sound like they're pulling stuff out of their ass.
Sometimes I do that. In my (partial) defense, some questions are not covered by dictionaries or style guides or corpus searches. Sometimes all there is is personal introspection (and linguistic science tries to build up beyond that). Other times...now I'm just being defensive.
ELU ain't Stack Overflow. You can't compile a sentence to see if it works. A sentence sounds good or not at all or somewhere in between. Also, as a native speaker, you don't have to have a lot of linguistic or self knowledge to tell if some grammar is weird or to come up with a vocab item but you certainly need both to explain well why the grammar or vocab might be appropriate. That is, non-experts can have something useful to add, they just may not be good at backing it up.
But your real question is about how you can trust what people say, or maybe where you're leaning is how you can trust a particular individual.
The structure of the SE software is intended to capture what the community (all the other visitors to the site) think about the questions and answers. Reputation is a numerical device that is simply the weight of votes for and against each individual's answers (and questions).
Reputation is a proxy for reliability. As a proxy, it is not perfect. People vote for things they like, how they're said, if the match what they already believe (confirmation bias), if they sound fancy (appeal to bombast), if they've said smart things before (appeal to authority, which may be a fallacy or not), etc, etc).
There are lots of visitors to ELU. Most just read. Some stay and ask questions. Some stay and answer. There's no guarantee of authority, just interest. There is a small handful of experts, actual professors of linguistics, people who are writers/editors. But for the most part, it's just regular people.
Sometimes people give their opinion stated as fact. Some people just cut and paste from Wiktionary or some site they googled, both of which have their own difficulties with reliability.
Surely having a high rep is no guarantee of authoritative knowledge. There are a handful of non-native speakers here who have gained a lot of rep points purely because their questions are so interesting. They are self-acknowledged learners of English and don't consider themselves reliable about English patterns. But they have a high rep. (there are also some non-native speakers who I would trust about their intuition on English, well, maybe almost as much as my own).
There are some suggested guidelines for writing answers which can make things more reliable. Citations from authoritative dictionaries (like OED or MW), examples of text found, corpus searches (from COCA or NGrams), all having their own qualifications and difficulties.
Yeah, I often see answers that seem like they pulled them out of nowhere or made up as plausible but there is no basis in fact. I find that annoying, and downvote if it is egregious or comment to ask for elaboration. The SE software is set up to deal with this kind of problem. That is, sure, don't trust the first thing you see and question the answer. This ain't as reliable as the Encyclopedia Britannica, but it also isn't as bad as Yahoo Answers (actually you can rely on YA to be not expert advice).
Of course, I usually fail to notice problems if I agree with the answer.
I realize that this is not a very satisfying answer. I'm not really saying either here are the desiderata for a user or a particular answer (because the reason you're asking this question is because rep is not very reliable), or on the other side that there are no good measures and it's the wild west and anything goes. Probably the best is a little experience here, seeing which persons write well, the form of which kinds of answers are trustworthy. Reputation points are a reasonable quick first approximation.
I think the moral is that ELU is an attempt to be as reliable as possible given that it is crowd-sourced by a number of people with varying levels of knowledge.