BUT NO! THERE'S A REASON FOR STYLE 1
You don't want to pay for the WRAP if the branch is not taken.
Unfortunately, given the nature of soft-quoting at the moment, that requires:
if blah blah blah @(wrap [
x: ...
y: ...
])
I've gone back and forth on the issue of whether group branches run unconditionally.
But looking at this concrete case now, I really do not like the @ being necessary there. Unfortunately, it's not as easy as saying "take the branch as a hard literal" because doing so ruins the tiebreaking that lets you put lambdas in branch slots.
Going to have to sleep on it. But looks like soft literal branches are safe.