Where it says that? Its simply that Python releases features in yearly cycles and thats what was completed for release.

Idea is to let people experiment with no-GIL to see what it breaks while maintainers and outside contractors improve the performance in future versions.

No that was not the idea. The feature went in under the assumption that the single thread slowdown would be offset by other minor speed improvements.

That was literally the official reason why it was accepted. Now we have slowdowns ranging from 20-50% compared to Python 3.9.

What outside contractors would fix the issue? The Python ruling class has chased away most people who actually have a clue about the Python C-API, which are now replaced by people pretending to understand the C-API and generating billable hours.