As I already said I am not in favor of asking the miners to use some conventions as for choosing one of several competitive blocks, yet I would like to note that apart from the suggested convention to choose the block at random there is a deterministic approach that is almost as good (meaning, it is not god at all):

We ask all nodes to propagate all competitive valid blocks. We ask the miner who receives k valid blocks B_1, B_2, …, B_k, to calculate the xor (or sum modulus 2^128) of lh_1, lh_2, …, lh_k and get a 128 bit random number R, where lh_j is the 128 lower bits of the hash of the block B_j (meaning, the bits that doesn’t affect the validity of the proof of effort). Then the miner should take the block whose index is R modulus k.
 
Despite being deterministic, this convention is having a similar effect as the convention asking miners to choose at random one of the blocks. If the attacker choose to keep her new mined block secret and release it only when the new honest block is published, since the lower 128 bits of the not-existing yet honest block are random, her winning probability is 1/2. Moreover, when the attacker is trying to discard a block that is already existing (i.e. replace the currently last block of the chain by on of hers), then although the lower 128 bits of the existing honest blocks are known to the attacker, the lower 128 bits of her own blocks are currently random, thus the winning probability is still 1/2.

Yet this convention is a bit worse than the randomly-choosing possibility:  in this convention the attacker knows immediately whether her block win or lose the race, instead of knowing this only when the new block on top of either of the competitive blocks is published. This knowledge can be used to improve the attack in case the attacker is secretly keeping yet another block on top of the recently released block, and can make a cleverer decision as for immediately releasing the next block or not. This exactly is the difference between the "st_m" family of strategies and the "sst_m" family, presented in my paper - http://eprint.iacr.org/2013/868.pdf.

Lear