This change caused a regression or two captured in issue36274. Essentially, by blocking invalid requests, it's now not possible for a system intentionally to generate invalid requests for testing purposes. As these point releases of Python start making it into the wild, the impact of this change will likely increase.
I think this patch was applied at too low a level. That is, instead of protecting the user inputs, the change protects the programmer's inputs.
I mention this here so those interested can follow the mitigation work happening in issue36274.