Supporting (or not) invalid identifiers in **kwargs

[guido (Guido van Rossum)]

@jeff5

This last one brings down the interpreter (idle in 3.8.10).

Is that a traceback or a segfault? If the latter, can you repro on main or at least 3.11?