Bastien writes:
>> I can understand the first question, but why the second?
>
> A mystery to me too, perhaps either Eric or Nicolas may explain.

That's because lob calls get wrapped internally in an anonymous
emacs-lisp source block that then feeds through the result from the
actual call as elisp.  The attached patch should suppress the
confirmation for the wrapper call.  To the best of my knowledge nothing
dangerous can happen with that evaluation and all confirmations for the
call stack down from there have already taken place according to the
users' setup.

Since the test suite doesn't really test confirmation at all, it would
be nice if someone with a good variety of Babel calls could check if the
number of confirmations stay the same with and without the patch.  I
don't think I have test cases that cover all possible combinations.