I investigated a bit more, and found that @a[*-1..*-1] just gives 4, and
f(|4) results in the same error message. So while the current behaviour
is not very dwimmy, it is at least somewhat understandable.
What's the opinion of the p6l people? Should |@a[1..1] work? even at the
cost of allowing |1 not to be an error?
Cheers,
Moritz