What criteria should be used when considering merge candidates?

Question

This question was prompted by the results of a normal dupe search on this question.

I discovered 2 possible duplicates as listed below and my first thought was that they should be merged.

Duplicate 1 Duplicate 2

However after close examination it's pretty clear that the answers in Duplicate 1 while having historical value are out of date. Duplicate 2 has more current answers, although the accepted answer refers directly to an answer to Duplicate 3?

So what should be done to cleanup this rather confusing mess in order to be most effective in assisting current users? What criteria should we be using to determine merge candidates?

It's not my intent to waste moderator time by suggesting merges that don't make sense but I do feel it's important that the end user be able to locate current answers that work.

Some of the things I think are worthy of discussion in this regard are as follows:

Age of the Answers: Are any of them currently on topic? I.E. Should Q & A's regarding EOL versions be included or skipped?

Referal: If answers in one question refer to answers in another question (or questions) is that something to be considered when determining a merge request?

Bad Plan: Is it just a bad plan to go through the merging process at a certain point and better to try to develop a canonical answer and simply mark all the merge candidates as a duplicate of that?

Answer 1

A merge more-or-less destroys one question, and groups answers from both under the master question.

Therefore my view is, a question is a good merge candidate if:

• it is a duplicate of a question which is better* or identical and itself (the question) adds nothing useful to the better one but
• both of the questions have good, useful answers, which are different from each other.

Don't merge if:

• duplicate posts have the same answer(s): there is no benefit in merging them.
• the questions closed as dupes are not exactly the same and both have useful irreducible details (in the question itself), then I think it is better not to merge them, because the questions themselves have value.

---

*better is subjective, but you may consider overall quality of the post itself (formatting, clarity, appropriate level of detail, etc), relevance to current versions, number of upvotes, number of views, number and quality of answers, whether an answer is accepted (if a post is merged, if an answer to the merged (destroyed) question is accepted, the writer of the post will lose their accept and 15 points) and so on.

Answer 2

Age Of the Answers: If the questions are substantially the same, on-topic and cover currently supported versions/or apparently timeless issues and there are answers of current value in both questions, by all means propose a merge and let those who have the capability merge them in order to improve the usefulness of the site if they see fit. Don't be disturbed if your request goes unfulfilled however.

Referal: If an answer to a more complex question refers to another answer as a step toward resolution then I would not consider that as a merge candidate.

Bad Plan: As I have yet to be faced with the task of merging questions I can`t answer this with a great deal of confidence, however I do believe that canonical Q & A's can be a highly effective way of providing quality answers to common questions in an efficient manner.

Edit: a review of Moderator Tools - Make merging questions a little easier seems to indicate that merging is not only a bit convoluted, but in most cases a really bad idea with many details to consider. I take this to mean that unless the candidates are identical in virtually all respects (including specific details in the answers) it's probably better to just mark as a dupe.