So, forgot what I was saying about mutexes; I forgot – there’s no easy way you can simply stop/start threads.
I have finally (been busy) duplicated the problem/solution for the freelist popping test bug. That is now release 6, known issue 6. The fix is to change the type of a variable in the test from unsigned int to atom_t.
I was looking again at SMR, via hazard pointers. I see a publically available C++ library using this, and I think JSR 166 uses it, too. So, I’m going to implement.