Because migration is different than actually closing a question, I think that the rules for close reasons that would result in migrations ought to be changed. Specifically, if a user votes to close their own question as "belongs on...", then the question should be migrated immediately or should, at most, require one other user to agree. Since the OP could have originally asked the question in the other forum, see little danger in allowing this and it would allow questions to be migrated to the appropriate forum more quickly. It is unlikely that a user would choose this option unless they realized later -- perhaps after seeing others voting to migrate the question -- that it was inappropriate for the original forum.
I think this is a fantastic idea. However, we should have a check/balance on this. Once the question is migrated by the OP (if that is the reason for the migration), they shouldn't be allowed to migrate it right away to another site or back to the original.
I don't know how often this would occur, but I'm sure that there is someone out there with enough time to simply migrate their questions all day for the heck of it...
I often find that the question asker is the least well equipped person to answer this question:
Which site does this question belong on?
Allowing one-user migration, or even two-user migration, is far too dangerous.
As Arjan suggests, we also need to let OP know what's going on with their question. Not everybody can see close votes distribution (you need 3000 rep) and 'belongs to' comments can be quite confusing.
1) What if we use standard notification on the top (like with badges, faq, etc)?
Somebody suggested that your question will be better answered at superuser.com. Click here to migrate it right now.
That's just a sample, but you get the idea.
2) Alternatively, there could be a comment posted.
Possibly belongs to serverfault.com. If you're the question author, you can close and migrate it right now.
Although, wording here seems clumsy.
3) Even better. Like now we have "migrated from" in the bottom, author could see something similar:
Migration proposed to
superuser.com x3 fluffykittens.com x1
Click to finalise migration.
<---- later can be replaced with "Please do not post duplicates, etc." if user has not enough reputation.
One point to consider is that if the user hasn't created an account on the target site yet, this would either result in a nonregistered user asking a question on that site, or the action being blocked while the user is shunted through the registration process.
Do we care? I don't know. But it's a complication that we should consider if we advocate a rule like this. Setting up an associated account on another StackExchange site is pretty quick and easy (as we on meta have all found out), but it's another layer that's going to be added to the process, both for the user, and for the SO developers to think about.
Note that I'm not necessarily opposed to doing this, I'm just saying that this part of the migration should be taken into account, and made as easy as possible if implemented.
I don't think that migrating a question is not different from closing a question; as a matter of fact, migrating a question also closes it on the site where it has been asked.
It doesn't seem to me that there are many users who votes to close their own questions, or migrating them to another site. I fear that, if users could migrate their own questions with a single vote, there would be some of them that would quickly migrate them to avoid somebody would down-vote the question, or they would migrate it right after a negative score, which would be reset in the site where the question is migrated (as far as I know).
I would prefer if the OP could cause the migration after somebody else voted to migrate, but I still don't see the hurry to migrate a question. There are questions that are closed from five users, and then re-opened from five different users.
What would happen if the question can be migrated because none of the tags is also used in the destination site? Should the question be immediately closed, left as if nobody voted to close it, or should the question require the number of users normally needed to close a question?