> Basically, let pp_next be a SVOP instead of a PVOP in that case.

Yes, I was looking into this this morning.

> But is it worth the overhead of yet another special case (and another
> test in pp_next) ? (just asking).

Probably not.