The primary axis of comparison I want to consider here is how technical the subject matter is. While this is not a binary, it influences necessarily binary policy choices, and is in my mind by far the most important factor in those decisions.

Curiously enough, the current footer does a pretty good job of sorting the sites into technical sites at the top and non-technical sites at the bottom (notwithstanding the meta sites - Meta itself, and the Proposals site - which have their own considerations and which I consider out of scope here). The sites I'd expect to play by noticeably different rules, when a choice needs to be made, are the first seven - Software Development, Electrical Engineering, Mathematics, Power Users, Linux Systems, Code Golf and Physics - versus the others (although the latter disciplines do have some varying degree of technical precision to them).

To illustrate: almost every question worth asking about Software Development is either a precise how question - how to do some specific, concrete task - or a precise why question - understanding a surprising, but reproducible result from an isolated example. Basically anything else has rapidly diminishing value - no matter how interesting it might be to experts. For example, if historical questions like "when was X implemented?" are asked at all, they are better shuffled off to the side.

In particular, debugging a thorny problem with inter-related problems might be engaging, but the result is not reusable for others. Even the best-asked question of this sort is of a fundamentally different kind. In the long run, everyone's time is best spent primarily on "canonical", "reference" material that isolates single issues that can be easily described. There are two reasons for this: because decomposing problems into those issues is supposed to be a fundamental skill; and because the resulting questions are more easily answered directly, more easily phrased, more easily given accurate titles, and far more easily searched for.

On the other hand, consider for example Cooking. Most recipes are just not that delicate, and even for the ones that are (say, to make a souffle), one would hardly speak of "debugging" the process of making the food. Natural ingredients vary in consistency; taste is subjective; there is a fair amount of tolerance in cooking times and temperatures; and hardly anything interesting will come out quite the same every time. Suggestions about how to get a desired quality in the final product might be much more open-ended.

As such, while there are a few things that one might identify as reference material (perhaps about food safety, or explanations of basic knife skills...), most of what needs to be said about cooking is far more individualized.

My basic idea here is that the less technical a site is, the less the conflict described in the OP exists. If the subject is a pure art form, a matter of historical interest, etc. then answering questions already inherently is producing quality, to the extent that it's possible. If I want to know why my recipe turned out poorly, it's relatively less likely that this can be answered by generalizing from a category of common mistakes. Or else, it's much more likely that there's more than one thing I could have done better - and it's unreasonable to expect me to try and take apart the process, check individual results etc. (Especially since trying to do so could involve tasting unsafe raw ingredients!)

But if the subject is a craft that uses reliable materials and aims for reproducible results; if it's something where new ideas are designed rather than inspired; or if it's something entirely abstract - then the opposite applies. On a hypothetical Furniture site, for example, it would be unreasonable to post a picture of my broken chair and ask why it failed to support an obese person. I should instead be expected to consider individual joints in my design (and the techniques used to fasten them), ensure that they are as strong as I expect individually, and then perhaps I can ask a question about why the whole system is weaker than expected (and maybe get an answer that uses engineering software on a computer to do a detailed model simulation).

So, in my view, it's in the most technical areas where the problem emerges most strongly. (And that's why it's been so noticeable on Stack Exchange: Stack Overflow absolutely dominates the entire rest of the network, and even "the rest of the network" is dominated by spin-offs from Stack Overflow.)