Just to reiterate some of the points previously said here:
Continental crust fragments that collided are sticky. This results in mountain belts that essentially "glue" the two crusts together.
Subduction of a plate that contains both oceanic and continental crusts underneath continental crust will inevitably lead the the complete loss of the oceanic crust, and the gluing of the continental crust to the other one. India and the Himalayas are an example of something like this with all oceanic crust lost to the mantle, and the Mediterranean is an example of an almost completely subducted oceanic crust (aka the Tethys), with the formation of a mountain belt (Alps) and soon have two continental crusts glued together (Europe and Africa).
it could also be subducting AWAY from an adjacent plate also containing continental crust
"Away" on a sphere means "towards" in the other direction. It's all relative here.
or at least lower the probability of all continental crust being
annealed at one time
They don't have to be annealed at one time. They just have to be annealed for long enough, without rifting, so all end up in one place.
is how all these pieces land together again & again so improbably
Improbably frequently? The last one existed about 300 million years ago, and the next one will form only in about 200 millions (more or less, depending who you're asking). In geological history, there are only two confirmed supercontinents, Pangea and Rodinia. Anything earlier than that is pure speculation (although educated speculation). I wouldn't call "twice" in 4.5 billion years "frequent".
I can also demonstrate by example. Today, there is only one rift that is properly breaking up a continent. This is the East African rift, and the Red Sea is a nascent ocean that is spreading now. But, there are more than one collision zones. The Australian plate is moving northwards, and soon will stick to SE Asia. India is already stuck there. Asian plate is stuck to European plate (Urals in Russia are the suture zone). Arabian plate and African plate are moving northwards as well (Alps, Zagros, all those mountain belts are the suture zones). The Pacific ocean used to be a whole lot bigger (Panthalassa Ocean). One oceanic plate is already underneath North America. The subduction rate of the Pacific Plate is the fastest of all subduction zones, and in not too long ago America (particularly North) will stick to Asia. The Atlantic Ocean has no subduction zones, so it will grow to become the new Panthalassa Ocean, with most other continents forming one single supercontinent.
the break-up is of no particular relevance, and seems to be straight-forward - plate movement under the cratons
Actually no. The "cratons" are already a single plate, glued together from several other plates. The break-up usually occurs because of
build-up of mantle heat under the insulating continental lid, will ultimately cause rifts to begin and start to form young oceanic crust (Arkenstein XII from the comments).