I was taking a look at some of my results, and I noticed some "result inconclusive" wingmen with obviously bad results (1.00 CPU seconds and only a few seconds of runtime.)
These were machines that were doing this on every workunit, and since each one is only running for a few seconds, they're doing THOUSANDS of them. They all seem to be Macs running Darwin.
Of course, when paired up with a valid result, they don't match, but when paired up with each other, they do.
So it appears that there's results in the PG database where two obviously invalid results are being marked as correct.
Here is one example.
My lucky number is 75898524288+1