Functions that coerce under some functor aren't very useful anymore.

[natefaubion]

Or at least, they are only useful if the functor is known at the call-site, otherwise you will incur a Coercible constraint. Previously just having the Functor instance was enough. Code that used this in a polymorphic setting will no longer compile. I'm not sure this is "wrong", but I'm not entirely sure those effects were anticipated since it's not really referenced in the release notes.