These kinds of incidents and errors look just like things programs written for Computer Science 101 classes would produce. This isn't some minor rounding error anymore - if true, someone has messed this up big time!
Have you seen the
BitGo integer overflow error? (BitGo gave to a lucky user a javascript tool to recover BTC from an old-format wallet. The tool had
an integer overflow bug that would send at most 2
32 - 1 satoshi (~40 BTC) to the destination output, leaving everything else as transaction fee. Fortunately the miner was nice and returned the ~80 BTC fee to the user.)
Yeah, heard about that one. Overflow bugs are always very weird things to watch. It totally makes sense for such a bug to be undiscovered in a recovery tool or something like that, but it's not as if the trading engine of Finex has only been live for a day by now. Still, I'd be very interested in seeing what actually happened here (if anything happened, actually!)