> Are the benchmark results that you posted above for removing the free list completely or for the one-element free list as in PR 14232?

Yes.  I see +3% overhead in several benchmarks and I want to see how one free obj save them.

> it's worse than no free list because you still have the overhead of dealing with the one object.

Yes.  But it still helps some common simple cases.
(e.g. PyObject_CallMethod() (not CallMethodObjArg))

> And it's worse than a free list of 256 because it won't for nested calls.

256 free list may bother 256 pools are reused.  But one free obj keeps only one pool.  And we can remove the overhead of linked list too.

Of course, I prefer a simple code.  But let's wait for benchmark result.