Some percentage of each grant should be set aside, not to those conducting the study, but to those who follow up on the study, with some award set aside for the first invalidation of its results.

I may be missing something, but... If there is a price for disproving a study that is a percent of the cost of the original study, then isn't that just making it lower payoff to cheat on the original study, and higher payoff to cheat on the disproving study?

That is, if I don't like the results that "Product X is ineffective", and I am willing to fund the study to disprove that claim, isn't it likely that I can more easily find a willing-to-fudge research team (because they are going to potentially get a bonus from the original grant's invalidation bonus)?

My understanding of this would be that that the original grant would be split something like 90%/10% (grant/invalidation bonus), and the second grant 90%/10% (grant/invalidation bonus) + 10% of the previous grant (if the original study is invalidated).