Let me preface this with, I'm not exactly sure how bitcoind deals with receiving two equivalent blocks (two different block 213s for example). Obviously, once it receives a block 214 it'll keep the 213 referenced by its 214. But until then, from my research (which may be incomplete, but I can't find much about this) it seems that it keeps the block it received first. This, however, is a breach of the bitcoin protocol.

The bitcoin protocol specifies that in the case of a blockchain fork the winning chain is the one with the most work in it. Clearly, the first block received is not necessarily the one with the most work - the block with the most work would be the block with the lowest hash. For all neutral miners (miners who didn't mine either version block 213) and non-mining nodes, replacing a block is a normal occurrence that doesn't cost anything. Therefore, following the protocol and replacing the accepted block by the block with the lowest hash seems like an intuitive thing to do. It is also a useful thing to do - it immediately ends the current situation where it is more profitable to mine blocks with less transactions, unless the transactions pay a much higher fee than otherwise would not be necessary. I see no reason why bitcoin core does not follow one of the most basic rules of the protocol, especially when not following it has a cost in network utility.

Yes, I do know that Gavin's O(1) block propagation time solves the problem too, however these solutions can coexist. And implementing replace-by-lowest-hash would be extremely simple, compared to the O(1) propagation time solution.

EDIT:
To avoid bumping this, I'm just going to edit this post.
I disagree with some of Shorena's specifics, but I can't refute the general point which is that it would open up some possibilities for attack that currently aren't possible.