> Amaury, if you decide to go forward with this, please clean-up the 
> patches to eliminate common subexpressions. 
I already considered this, but generated machine code was identical, so
I chose the more readable code.

> Wonder if there is an alternate solution that keeps the next slot 
> filled with something that raises an Attribute error.

That's an interesting idea, and not difficult to implement, see attached
patch. The penalty here is an extra comparison in PyIter_Check(). And no
change in the event loop...
Armin's crashers now correctly raise a TypeError (and the final patch
should convert them as unit tests)