As another moderator pro tem on Crypto.SE I second everything Jeff and Paulo have said so far. I'd also like to make the following observations:
I think both Math.SE and Crypto have a large amount of overlap. Clearly if you're going to discuss cryptography, chances are you're using some Mathematics somewhere. The edges are more blurry when there is a lot of Mathematics involved. That said, there are different considerations in cryptography than necessarily apply in Mathematics, although the two inform each other. I guess from our point of view, a differentiating factor might be:
Would this benefit from the point of view of a Cryptographer who knows some Maths, or a Mathematician who knows some Cryptography?
Second point - I have a feeling that in general some people see the Area51 process/StackExchange sites thing as a chance to define an ontology. I personally think whilst ontologies are interesting, too much discussion and too many attempts to define a totally precise way of determining what question belongs where detracts from the overall goal of actually getting a question answered by the right people.
So this is an open invitation not to do that between Math.SE and Crypto.SE. Life really is too short. If users (including moderators) want to hop into Crypto Chat or elsewhere and ask us what we think of a given question, great, excellent. That sort of discussion is exactly what should happen. I shall also try to be present on Maths chat too. The way I see it, the two sites technically feed into each other. Anyone wanting to "get into" cryptography who doesn't necessarily have all of the background Maths might well be better asking here for some of their questions, for example. Also. we would absolutely welcome some maths users wandering over and providing feedback on our more mathematical topics. For example, Thomas's answer on my question is pretty good, but I'm always open to more involved feedback.
In summary: what should be migrated should be discussed - a set of hard and fast rules that apply to every possible scenario do not exist. This is why we have human exception handlers and voting - to make the judgement call.
I would also re-iterate Paulo's call for talking to us ASAP regarding migration, or Gilles's feature request for vote resets, for the reasons Paulo states.