Hi Marco

let's start with the bottom of your mail:

 >
 > I hope it's ok to post my thoughts here.
 >

that's the purpose (even though not that much used)

>> The team intend to modify the existing \<thing>_case:nnn functions,
>> renaming the last argument as 'F'. This reflects the fact that this
>> final argument is used only when the test case (<thing>-dependent) is
>> logically false, and follows the approach used elsewhere.

> I think it's not a good decision. The argument specifier F implies for
> me that there is also a true part (T). So I would prefer to use
> \<thing>_case:nnn.

As Joseph said, there are other similar cases. And in my view if you 
request a selection (and you fail to match) then the else is a natural F

In fact one could think of also having

\<thing>_case:nnTF

where the T branch is selected if you have a match (in addition to 
executing any match code from the "n" argument). I personally rather 
liked that I idea, as it fits well with other places like \prop_get:NnN(TF)

cheers
frank