Well no because that assumes there are no other transactions competing for space in the block. Already today in order to get priority processing many users pay > 0.0005 BTC. An attacker attempting to launch such an attack would simply find that other users raise fees as they see their confirmations slow down. A smaller and smaller % of attackers fee are put into the block. Worse many blocks would go over the 50% mark and the fees required for inclusion of low priority transactions would skyrocket. The network is well protected against a spam attack. If necessary the mandatory fees (which AND I CAN'T STRESS THIS ENOUGH only affect low priority txs) can be raised by a consensus of miners. It has been lowered in the past. Even without raising the fee the USD cost will rise in relation to the USD:BTC exchange rate.
The attacker would be just spamming, not trying to push out normal transactions from blocks. If he sends 1000 spam transactions per block at 0.0005 BTC/tx, then according to the described mechanism they will be added to the blocks so the blocks are filled up to 50% of the limit. If some block happens to be this size already, then none of the spam tx is included, however, it is still a big block.
Okay. Assuming that the miners don't like spam either, I agree it is unlikely that any big spam would go either unnoticed, or noticed and still packed into blocks by miners.
I think the full nodes are very important to the network, and their significant decrease, due to e.g. spammed block-chain, would be harmful. But it's just my opinion. Or maybe the users would quickly migrate to the light-weight clients, who will quickly attach to the remaining full nodes, who will quickly start charging for their services, and problem solved.
It's a wholly interesting project!