More blue-sky LF thinking today; I think I can get away with polymorphic lists that actually support hd and tl eliminations by using the same sort of by-need η-expansion as found in Nanevski-Morrisett-Birkedal-style polymorphism.