I recently came across this story. I'd heard about it at a NSF panel but hadn't gotten all the details. This person (who plead guilty and has not yet been sentenced that I can see) did at least two bad things. First, and obviously illegal, he had a NIH award in which there was supposed to be a significant subcontract to another institution, and instead he spent that money on something else (possibly even on personal stuff). It's actually amazing to me that he didn't get caught earlier on that by his institution's research office. As chronically understaffed and overworked as ours is, they are zealous about making sure that subcontracts and reporting are properly handled, so I don't see how something like not passing along $500K could happen.
Second, and potentially trickier, after getting an award from the NSF, he applied for a grant from the DOE's ARPA-E for basically the same work, without telling either the DOE or NSF about the overlap. That's also illegal, though I suspect it's more common simply because there are shades of grey possible here. Research projects can have overlap - particularly if a PI has a particular technique or tool that they've developed and want to push in many directions - the question is, how much commonality is too much? This particular case was egregious. Still, after reviewing some grants recently for a few places, I want to encourage my junior colleagues to take these issues seriously. When you review a proposal and realize that you've actually already reviewed something with lots of overlap before from the same PI, and it was funded, yet the PI claims there is no overlap in the programs, it's not a good situation for anyone.