i haven't in part because i'm lazy and hoping someone else will do a good, controlled review and i can just read it!
random thoughts:
1) i think you need a really good, static test scene with lots of nice gradients of light and shadow. that is to say, _gradual_ drop offs from bright light to deep shadow. color and pattern as well as you can get color impact and patterns/textures may be affected by compression.
2) i'm guessing performance will vary by iso and by exposure. so an optimal test would systematically sample different isos and different exposures and combinations thereof.
3) i think you want to focus on examining the highlights and dark shadow areas for detail as those are most likely to show impact.