Often obvious recursive algorithms aren't well-optimized by GHC. This might be the kind of "bend over backwards" coding you're talking about, though.

The way you can escape from this performance trap is to (as usual) write explicit state passing instead of direct recursion. GHC can inline and fuse these away much more nicely and often arrives at a very tight, C-like inner loop.

I'm not sure if I call this bending over backwards though since Haskell style has always been to avoid explicit recursion and this kind of state-passing transform can be hidden behind the same set of combinators you're used to.