After a little introspection, I've realized that Minor already has a
SRFI-50-like interface implemented --- complete with critical section
markers, direct references to heap objects, etc.  But it's strictly
internal; it's what I built the JNI-like interface on top of.

I deliberately chose not to expose that interface, because I felt it
would be too error-prone, my users (should there ever be any) would
spend too much of their time struggling with it, and I would spend too
much time trying to explain it.

But when I write up the specification for how to write machine code
that interacts properly with the Minor allocator, GC, and thread
support, a good portion of that stuff will end up in the public
anyway, to provide header files for people generating machine code
directly.

Hmm.