Comment by _flux

Wouldn't ref-counting CPython already know that a has a single reference, allowing this optimization without needing any particular smart JIT?