I think the correspondence between pattern matching and initializers in Scala is meant to be achieved at the grammar level. I did not see the symmetry at the bytecode level between unapply, unapplySeq and apply.
This proposal does not affect the correspondence at the grammar level. In fact, the original intention of this proposal is to keep the same translation rule for both pattern matching and initializers in name based XML literals when optimizing the performance of that proposal. Explicit currying cannot achieve the goal.