Edward Kmett just introduced one in another thread. Simplifying, it would be this:

For all x, y, f:

do { x' <- x ; y' <- y ; return (f x' y') }
-->
f <$> x <*> y

This is a great example, because (1) it reduces clutter and "temporary" names and (2) requires significant background knowledge on monads and idioms. We can also generalize this to functions f with increasing arity (f <$> x <*> y <*> z, etc.). Beginners would not get this, but once you know this rule, it can greatly improve your coding style. Similarly with liftM2, liftM3, etc. as mentioned by Stephen.

Any other useful tidbits to share?

Sean