Yeah, I'm aware of these two commits (as one bisection on anderson landed there) but they got merged only in 5.1-rc1. So they aren't a reason for this regression.