I dunno, I kind of disagree? Like, just because you have dependent types, doesn't mean you verify stuff all the time? Not disagreeing that this stuff is hard - I just think it would be a shame if Haskell turned pushing the research side of things.
It also seems like working on dependent Haskell is forcing them to clean up a bunch of tech debt in GHC, which does seem like a pretty nice thing as well.