I wish more people did mutation testing.
But, right now, MuCheck for #Haskell makes it really obnoxious because everything to be tested AND all the tests have to be in the same module.
That said, I am _entirely_ uninterested in trying to parse the GHC source language in order to mutate it myself. I wonder how awkward lib-ghc-parser is these days. (Only being able to mutate Haskell-by-the-Report would be near useless in practice.)
It really feels like the part to be mutated can be determined by the coverage HPC is already generating, and True / False mutants can be generated easily enough from the places HPC already flags as always false / always true. Running the test suite again on the mutated code is a bit awkward, but tighter integration with cabal might make it easier.
Using Quickcheck (Arbitrary) or Smallcheck (Serial) could generate mutants for any covered (non-Bool) subexpresison that has the right instance.
I've got far more "urgent / important" stuff to work on, tho.