The way I see it, it should not be further reducible. I'm thinking
λz.zq is like
lambda z: z(q) # Python, not lambda calc
and you do not know z nor q so it is not further reducible.
The below derivation (where I am wondering about the last steps) is from the new Haskell book: