Has there been any analysis of an "empty block attack"? Imagine what would happen if an inordinate number of empty blocks were solved and submitted in succession? Bitcoin would effectively cease functioning.
Someone (or a group of people working in concert) who can solve and submit an inordinate number of blocks (of any type) in succession can do bad things. This is a well-known problem for which there is no possible solution (at least no solution within the concept of having a decentralized proof of work blockchain).
Fortunately, in order to submit an inordinate number of blocks in succession with any significant probability, you have to have a lot of "mining power" (51% would definitely do it; maybe 33 1/3% under the right situations; probably somewhere in between the two under less "ideal" situations).
Seems to me the empty block thing is a huge oversight, and that empty blocks should not be valid and the network should reject them.
It's trivial to add one (or any number of) useless spam transactions to a block. If you're trying to save network bandwidth you could even calculate the spam transactions based on a formula, so you don't have to send the actual transaction content over the network. So rejecting empty blocks wouldn't add any security, and would really just add complications to the client. It could also potentially make Sybil attacks slightly easier (all you have to do to stop a miner from mining is stop them from getting new transactions, you don't have to shut down their network connection entirely).
You could have a rule where every block must have at least a certain number of bitcoin-days destroyed. But that would have its plusses and minuses. Something for an altcoin to consider (especially if you added something fun like negative transaction fees for "high priority" transactions), but not something that's necessary for bitcoin.