I'm a novice C programmer, but I'm unsure about the safety of your
thread-safe collections description.
The "list" class uses a slightly different strategy than "dict", which I forgot about
when writing the design overview. List relies on the property that the backing array
of a given list can only grow (never shrink) [1]. This is different from upstream CPython.
Dict stores the capacity inside PyDictKeysObject (the backing array). The capacity
never changes, so if you have a valid pointer to the PyDictKeysObject you load the
correct capacity.
I've been meaning to change "list" to use the same strategy as "dict". I think that would
simplify the overall design and let "list" shrink the backing array again.