Make the simplifier less thunky
Currently the simplifier contains a fair number of thunk allocation. With no doubt many of them being evaluated in short order.
It would be good to do a systematic pass over this code to decide which places in the simplifier make good use of laziness. And which just delay evaluation when we shouldn't.
But in the meantime I created !4808 (closed) which stomps out some of the more obvious thunks. The end resulting is 3% fewer allocations by the simplifier when compiling Cabal with -O2.