Since their introduction, migrations have been largely well received, but still cause some problems, typically in the area of folks migrating questions to sites they didn't understand or migrating subpar questions.
One of the key guidelines for migrating questions (after "don't migrate crap") is "don't migrate old, answered questions".
We recently disabled migrations on questions older than 60 days, which goes a long way towards resolving problems with such migrations, but does leave two holes:
- New sites (like Mathematica) that form successful communities of their own, even if born out of a larger site like Stack Overflow.
- Old questions that haven't received answers but fell through the cracks of time.
The second is easy: if the question is off-topic on your site, vote to close it as such regardless of its age or quality. You can leave a comment on the post suggesting a better location. If the asker is still around and interested in the answer, they will be grateful and will repost the question in a better location to get their solutions.
The first, however, requires a bit more of a process. Let's say you're on a smaller site that has grown to the point where you want to start considering picking some canonical questions that have already been asked and answered elsewhere. Here's how you can get some (or all) of them migrated to your site:
Start on your meta. Build a list of questions that you think would benefit from being migrated to your site and collect some reasons along with each one. Don't just say "all questions in our tags!". That's neither going to happen nor is a good approach - do you really want to swarm your new community with old content?
This is also your opportunity to edit old questions to make them more in line with your scope and standards.
Once you have your list...
The questions will be vetted by the source community as well. The exact details here are still to be worked out, but the idea is that we would present these questions to the relevant subcommunities of the original site. For example, questions that are candidates for migration to Mathematica would be vetted by the regulars of mathematica on Stack Overflow.
Some form of discussion will ensue and by the end of it we will have the final list of questions to move over.
We will do a one-time mass migration of the questions that make the cut. This has some additional implications. For example, your front page on the smaller site will be more or less destroyed for a bit, so make each question count.
Once this migration happens, the topic is closed. There will be no more discussions, migrations, exceptions, etc. where old questions are concerned. You're free to still edit and migrate new questions if needed.
What happens if some of the questions you wanted don't get migrated? Just treat the original site like you would any external blog. If someone asks a similar question on your site and you're certain that you can't do better than the original version elsewhere, summarize it, give credit to the author, and link back to the original.
In my experience, there are definite reasons to sometimes migrate old questions to new sites, but those are rare and should remain exceptional. This system should hit on those cases, allow both communities to have a voice, and miss ones where a migration would do more harm than good.