Benjamin, what "complexity" did you encounter when trying to use lib2to3
in your own work? Unless there's a concrete use-case where the mere
existence of multiprocessing support (as opposed to actually enabling
that support) made a tangible project harder, I say we leave it in. I
can't imagine someone making the conscious choice to slow down their own
lib2to3-based refactoring tool by 2x; if those people appear later,
we'll deal with their concerns at that time.

Nick: if you don't have time to make this work in 2.6, I can do so. The
patch will also need doc fixes and tests, though the core functionality
is fine.