look at this. 80K unconf tx's and 6.76TPS and yet gmax et al says there's no problem. all that wasted computational work done by full nodes for blocks that will never clear those unconf tx's. i told you long ago these guys will never even be able to recognize a problem even when it's staring them in the face and even after it's beat them over the head:
So the fix is to happily write these 80k spam transactions into the permanent ledger that we (full nodes) have to keep a copy of forever?
This is an order-of-magnitude estimate, but I think it's pretty convincing that it's both more expensive for the spammer and less expensive for a node/miner to store the spam than to reject it. When the spam is rejected, it can be "re-spammed" without the attacker losing the TX fees. Once the spam is written, however, the attacker needs more funds to perform a new attack.
I think you have to multiply the cost with number of full nodes.
Why would the cost to a particular node depend on the number of nodes?