I reckon deriving answers from all source classes (primary/secondary/tertiary) is valid and each have their place on the site. Regardless of the source, the onus is always on the answer to actually demonstrate its value.
No matter how well-intentioned, limiting whole classes of sources has the problematic side effect of also limiting experts. If we forbid deriving answers from primary sources period, we by default also forbid answers from the very experts who are actually able to derive such rulings: I think we can all agree that we want that sort of expert active on our site. Just like we want experts who are well-versed in synthesizing answers based on secondary sources. Or experts who know when it's best to rely on tertiary sources. The important thing is, we want experts because they can answer questions.
On the other hand, if we as a site choose to treat all users as if they're laymen (e.g. wholly unauthorized to derive rulings from primary sources), then as a site that's exactly the level of expertise we're going to actually develop.
There's no reason we cannot have both laymen and experts on the same site. What is important is that all users, regardless of their level of expertise, are only writing answers that they are actually qualified to answer.
If someone writes an answer that you think should only ever be written by a certain class of expert, judge it according to what such expert answers should contain to be convincing. Be critical. Be harsh. Be whatever you think it takes to prove that this answer is able to support itself for what it's claiming to be. Since we have no way of proving who is who on the internet (and especially their credentials) voting is of utmost importance to ensure that the expert answers are written by actual experts.
Is deriving a fatwa based on the apparent meaning of a single hadith useful? Probably not; as mentioned in OP most if not all schools require a rather comprehensive level of academic background to even consider deriving rulings. Even when a fatwa from a reputable scholarly source appears to be deriving itself from a single hadith, it comes with the implied background of knowing not only the variant narrations of that hadith, any potential hadiths that conflict with it, any potential different interpretations of meaning, the context in which it was presented, different scholarly opinions as to the validity of the hadith in the context of the derived ruling, and the relative strengths of all the above.
Can all that implied background be sufficiently demonstrated in any post of reasonable length? I don't know; I'm not a mufti. But if you're certain that an answer is plain not useful without it, it should be downvoted for "This answer is not useful". If, somehow, the answer actually convinces you that it does speak with sufficient authority, then upvote it.
And even if an answer is basing itself wholly on secondary sources, the onus is still on the answer to demonstrate its usefulness. Does a century-old fatwa by a Saudi scholar have any bearing on modern-day Canada? Does this fatwa come from a scholar who is considered authoritative in his field? Does this fatwa represent a majority opinion, or is it a fringe interpretation? Again, I don't know; maybe it is and maybe it isn't. And again, it's up to the answer to demonstrate to everyone else that it's actually useful to the question asked rather than just some random fatwa he found in a book somewhere.
In other words, answers should be judged for what they are. If an answer appears to be deriving a novel fatwa, by its nature is basically claiming authority to derive such and it should be judged on the same level as you would judge an answer by someone with that actual authority. If the answer is unable to demonstrate the actual expertise to back up this claimed authority, it should be downvoted and downvoted harshly.
There's no reason we can't judge a layman harshly (as should be the case) if he tries to derive a ruling based on insufficient understanding while at the same time commending a scholar who derives one from the same sources; what matters is how the information (and the expertise it reflects) is presented.
