Tl;dr: not allowing data fabrication is some kind of lofty, unreasonable standard that doesn't allow science to be published.
Please.
Please.
Ed: to be clear, the one making the argument about publishing pressure in a sensible way is https://news.ycombinator.com/item?id=36881029
Observing that fixing the latter might eliminate most of the behavior doesn't mean you think the behavior's good.