> As a miner, you don't want to have an orphan, so you'll have to discount the chance of being orphaned on the big block chain against the potential benefits and costs of mining the other chain

But that cost is paid only once per fork.

>  if the fork gets orphaned and you mine in the small chain, even if you want big blocks, you end up with big block coins AND small coins!.

On the other hand, if the big fork is not orphaned, you get ONLY small coins.  If the difficulty is the same after the fork, and expected market prices of the two coins are Ps (small) and Pb (big), switching to the small chain may be worth only if 0.1x(Ps + Pb) + 0.9xPs > 0.9xPb, that is,  if Ps > 0.8 Pb.  If the general belief is that the small coins will die eventually, then they should die immediately; that is, Ps should be near zero.

By the way, note that miners cannot move their reward coins until they have 100 confirmations.  So the unfaithful miner would have to gamble that Ps will not crash before that.

This is also the reason why we don't see a 25% miner creating his own branch of the blockchain and selfishly mining it, ignoring the blocks of other miners. That would create an altcoin that is valid for the mainstream miners, so the system would evolve just as after a hard fork -- with MainCoin and AltCoin in place of BigCoin and SmallCoin.  The same reasoning would apply: if that miner is lucky and the AltCoin branch overtakes the MainCoin one, a reorg of MainCoin will happen the miner will get reward in both coins, otherwise he will get only the AltCoin rewards.

> I find it a bit confusing that a safer method for splitting chains isn't applied, one where the chains clearly diverge.

Yes, I think it would be better if changes to the protocol were done by "clean forks" instead of hard or soft forks. See [this earlier post of mine](https://np.reddit.com/r/bitcoin_uncensored/comments/3hj2sl/soft_fork_hard_fork/), last section.

> This lack of concern for "edge" cases is typical of what I've seen in the Classic team, and unfortunate

The Core team seems worse in this regard.  See the Fork of July fiasco.  That 6-block reorg was due to BIP66 being enabled with no grace period, almost ensuring that 5% of the miners would not be upgraded yet.