Thomas posted a patch yesterday that seems to take care of the callback deallocation problem I was having with OpenGL-ctypes. Unfortunately, it didn't move me very far forward. I'm now seeing a crash inside the GLUT mainloop, particularly during a call to glFlush().

Only thing I can think of at the moment that would cause the effect would be a pointer that was used to pass data getting deallocated before the call to glFlush prompted the operation to occur... but that's not happening. It's only happening with the OpenGLContext context, when I transfer the same sequence of OpenGL calls into a plain-Jane GLUT-only file everything's fine.