I cannot speak for Martin, but I think that what he meant is that if existential types have been removed, the likelihood that universal quantification of types makes it into the compiler is minimal.
I think this has to do with two things:
- The complexity that it would bring to the typechecker (even more than existential types, I think).
- Challenging implementation details, up until the point which the current compiler does not allow an easy implementation of universal quantification.
However, I think true polymorphic values would be useful. I don't know, maybe if someone steps up and offers an implementation, he can reconsider? I suggest you ask for a better explanation in the Dotty ticket, or even suggest looking into its feasibility if you're truly interested.