Unfortunately, I don't think such a study is runnable. How do you blind 'oil-sludge' against 'real food'?
That said, randomizable is perfectly doable: Rhinehart could flip a coin each week to decide whether to eat real food or Soylent. I suggested as much on his blog a while ago, but he never responded.