This isn't a matter of 'a majority of bitcoin miners refusing to increase the limit'. This isn't something the miners could pull off themselves even with perfect cooperation among them. Producing additional coins above those specified by the bitcoin protocol is visible. Everyone can see whether the number that happens to be hashed represents a bitcoin or some other thing the miner decided to make (a NotBitcoin). The task of the miner then is to find someone who is willing to buy his newly minted alternate currency for the same amount that a bitcoin sells for. That is difficult. Very difficult. Especially since the change is from something stable to something that is already known can be changed on a whim.

I should have been clearer about what I meant by increasing the number of bitcoins. I meant altering the bitcoin protocol so that either the per-block reward is increased or the number of blocks between the reward-halving is increased, or that the fraction by which the reward decreases is decreased. Changing the current per-block reward would be much more difficult because it wouldn't be accepted by the majority of clients. Changing the reward-halving rate or fraction would be easier because it could be implemented in the current clients to occur far enough in the future that the vast majority of clients would accept blocks with the altered protocol. If a sufficient number of miners began using the new protocol it would be difficult for exchanges and normal users to stay on the old protocol after the fork; the delay between blocks in the original-protocol chain would increase significantly until the difficulty adjusted and having a split networkd would make it very difficult for the minority supporters to do business with bitcoins. Perhaps the old original chain would continue and be revalued according to its new utility, kind of like gold and silver coins were kept around for quite a while after the Federal Reserve started issuing notes. Neither chain would have precisely the same value it did before, but they could diverge fairly smoothly and slowly if it was a small change in the protocol.

In other words; if enough people think a permanent finite supply of bitcoins is a bug, it will be treated like previous bugs in the block chain. It will be fixed in advance, with users and miners given enough time to update to the new clients. So long as at least one exchange and the majority of miners agree on a protocol change I think it's inevitable that it will happen.

This is true but an innaplicable and entirely misleading analogy. To try to express what it would mean in those terms it would be if there were federal reserve notes called "dollars" that can be redeemed for silver and an entirely different kind of note that someone tries to call "dollar" that cannot be redeemed for silver. Expecting people to assign just as much value to the latter as the former is difficult. Especially if there is no government there decreeing that it is law that everyone must accept NotDollars as legal tender.

But that's precisely what happened. A US Dollar(1792) is worth "three hundred and seventy-one grains and four sixteenth parts of a grain of pure, or four hundred and sixteen grains of standard silver". A US Dollar(2013) (Federal Reserve Note) is worth less. If you actually possess a US Dollar(1792) it is still worth 371.25 grains of silver by definition, but probably quite a bit more to a collector. I don't know the complete history of the Dollar, and its change in value didn't happen overnight, but it's probably mostly because of the length of time it took for that change to occur. I don't think a large change to the bitcoin protocol would be accepted today. But a planned change that would occur in 5 years and was expected to incur a minimal change of value at that point? Definitely.