So many potential checkpoints in this experiment seem to have been missed. As others have already pointed out, everything from not estimating the dosage result before performing the (inaccurate) calculation, the point that 30g of caffeine is a BIG pile of powder, the fact that it would be unpleasant for the testers to drink (and would have taken a lot of mixing, due to not readily dissolving), all should have triggered alarm bells with the students performing the test, yet they seem to have blundered on regardless.
When building anything, my father in law always says "measure twice, cut once". A similar philosophy here would have hopefully prevented this incident occurring in the first place.
Biting the hand that feeds IT
