I'd rather not mix the kernel scriptlets with this. Replacing the scriptlets and replacing regenerate-initrd-posttrans are two independent things. I think creating another subpackage is the right thing to do, even if I previously said the contrary. It's consistent with the way we've handled this previously.