As far as I know the only time wrap exploit that has ever been clearly identified and described is the original off-by-one bug identified by ArtForz in BTC and its clones.
Evan claims a KGW exploit was deployed against DRK.
I see no evidence that anyone on that thread has any idea what they are talking about with the possible exception of Cryddit. Nevertheless KGW looks like kind of a train wreck to me, and probably DGW too (but I haven't really looked that closely).
Clearly you see now the potential problem in Cryptonote with the 20% discard rule. It enables the secret chain to hide a bunch of blocks without causing a rise in difficulty.
No, it isn't clear. You can't really hide any blocks just by making them outliers because the outliers starting at the most recent end of the adjustment window (for example if you timestamped into the future) still have to slide through the middle before exiting the window on the other end. So you can only defer them from contributing to the adjustment for a little while, but eventually they do get counted (similar in effect to a window-based adjustment like Bitcoin). The outliers at the farthest-in-the-past end might be able to slide off without ever being counted, but even if you could figure out how to drop blocks there right away, that would only increase difficulty, not decrease it. There still might be a flaw, but we have to do better than that to graduate from FUD.
Thus the secret chain can end up the longest chain without needing 50% of the hashrate.
This can never happen because the chain length is sum of difficulty not block count, although with some probability you might have a slightly lower hash rate and still get lucky and win more than half of the (weighted) blocks, as usual.