No each thread does it’s own alloc/free. I reduce the locking to a minimum by using per-thread caching and memorisation. I had to implement that to make the code scale linearly with number of cores. Without per-thread caching it would only scale to about 3 cores. Now it scales very cleanly. Which is nice.

That's not enough. Imagine a circular linked list. GC looks at an outside pointer to item A in it. Now another thread removes A from the list and moves the outside pointer to the next item. After that, GC would see all items in the list not referenced by anything (apart from the cycle).

Yeah, I didn't say it was sufficient. I just said it's probably one of the things they do.