However, please be sure that the below answer actually answers OP's question before voting to close! If the question asks about something not covered below (other than the specific fields), the question should be left open until its answers are merged into this canonical question.
I do not have moderator powers and don't anticipate earning enough rep for them, so I'm hesitant to comment on the precise wording of the preface text.
But as I requested it, maybe some further explanation as to why would be useful.
As a totally anecdotal observation, this Stack closes questions more often compared to some others. I think this is good-- we have relatively low volume, so can afford the time to try to help askers find an answer to their question. The busier Stacks seem to rely more heavily on downvotes in contrast, which can be pretty frustrating for the asker since they don't necessarily get to know what the issue was.
The downside of our predisposition is that borderline questions are more likely to be closed on balance, which is final in that it stops answers from accruing and is still jolting to proactive askers. Once a question is closed as a duplicate, the burden of making an argument convincing enough to reopen it usually falls on the asker. But the close voters are more likely to be familiar with that context in the first place, so clearly should have more responsibility for knowing the answer base.
I suggested some form of warning for this specific canonical question because it seems to me there are a lot of related questions and answers in the genre that are within scope and unlikely to be consulted by general advice answers. The way this concern arose was straightforward: When I tried to think of a simple typology from which to construct answers to a canonical question, I had difficulties.
Cag made a compelling case, however, that we have a history over time of receiving essentially identical questions and producing essentially similar general-advice answers. A preface warning's content may seem obvious, but my thought was it represented a reasonable PSA in a likely-to-be-seen place that succinctly highlights this tension to flaggers and voters.
On a particular point:
We do not need more rules and special cases.
I agree with this and do not think the preface is either a rule or a special case. It is not a special case because all duplicate voters/flaggers should make the suggested consideration in every case. It is not a rule in the same sense that almost every moderation decision on Stack doesn't arise from an explicit rule but from some consensus amongst the users who have moderation powers. In particular, none of the diamond moderators suggest they will user "superpowers" to enforce it.
That being said, I wouldn't be bothering with these replies and suggestions if I didn't think they were useful. I hope Stack members will at least give them a thought if they interface with this particular genre of question.
The instructions are stating what should be obvious: don't hammer a question for being asked-and-answered unless it has actually been asked and answered.
It is understandably infuriating for the asker when they ask "how does X affect Y" and it gets closed as a duplicate of "how does Y work?", but the linked question does not address X, does not address any generalization of X, and does not explain why we're not addressing X.
Moreover, this is a problem for the site: if the canonical question should address X, but it doesn't, and we close all questions that ask about X, then there's no way for askers to learn about X or for answerers to realize that their expertise on X might be valuable.
I won't go line-by-line through your points, but I want to address the first two:
It is impractical to determine the contents of a long and changing community wiki for each potential duplicate.
Disagree completely. If you are going to close someone's question as a duplicate of another question, it is your responsibility to verify that it is actually a duplicate. It is not good enough to say "well, they're basically the same, it's too much work to make sure it actually contains an answer." No one is forcing you to exercise your close votes or reopen votes, but casting a duplicate vote means telling a human being "your post should be deleted because it's been asked-and-answered elsewhere." If you're going to do this, you should take the time to make sure that you are correct.
Voting to close a question as a duplicate of another question should be based on the content of the question, not the content of answers.
If the two questions are identical, then there is no issue: close away! The instructions don't state otherwise. As discussed above, the issue is where to draw the line when the questions are close but not identical (most often, the new question asks about a specific case of the canonical question).
[By the way, I did not invent these instructions out of the blue, they were specifically requested here].
Update, 6 mos later: I decided to reword these instructions to make them a bit more concise. The "question closure instructions" have now been edited into the main canonical question intro paragraph. This is just wordsmithing and does not imply any change to our closure procedures.
Here's a possible alternative wording:
Note to close voters: please consider this meta discussion before closing a question as a duplicate.
These instructions should be removed because:
- It is impractical to determine the contents of a long and changing community wiki for each potential duplicate.
- Voting to close a question as a duplicate of another question should be based on the content of the question, not the content of answers.
- Merging answers is impractical and undermines the answer voting process.
- Redoing close voting after merging answers is also impractical.
- If it is unclear if a question is a duplicate, the meta site provides a good mechanism for resolving any uncertainty.
- We do not need more rules and special cases.