Of course it's Luke and his "gut feeling" transaction validation algorithm. He has a history of trying to push filters for the transactions he personally dislikes or considers not being real in some sense. Be it satoshidice, OpenTimestamps, ordinals...
But let's take his claim at face value. Does the attacker gains cheaper blockspace by abusing this vulnerability, compared to regular txs of the same size? And if so, why are the miners including them in the block template?