So now, it's happened before. Care to back up your claim?
In 2011 (or 2010?) someone noticed a bug in the implementation of the multiply instruction, that did not properly guard against overflow. Exploiting that bug, the guy managed to creat an unspent transaction output with a humongous amount of bitcoins. In order to fix that bug, Gavin and other developers had to perform a hard fork, starting from some block before that buggy transaction. (You can easily find this story in the internet.)
So, for a while, there were two clones of bitcoin, starting off from that same block: one compliant to the old software, with the buggy OP_MUL and the gazillion BTC output, and one compliant with the new version, where the OP_MUL instruction was disabled. Most of the transactions that had been processed in the old chain after the fork were re-issued on the new chain, but some (inclunding that one with the gazillion output) had to be discarded. So the two chains were effectively two similar, but not identical, versions of bitcoin.
By common agreement, the old chain was abandoned and everybody switched to use the new chain, by upgrading their software. But anyone could have continued to mine the old branch (except that the difficulty would have to be lowered artificially, for a while, to allow a decent block rate). Any bitcoiner could have continued to move, spend, sell his old bitcoins, using the old software and the same private keys, independently of what he did with the new bitcoins in the new branch. Of course, the OP_MUL bug would have quickly made a big mess of that old branch.
But the point is that the the old chain was not "killed" by some mechanism in the protocol, it was just abandoned because no one had interest in it. If the bug had not been not so severe, perhaps some people would have continued to use it, for fun or even for profit. Anyway, while the two branches ran in parallel, the number of existing bitcoins was effectively doubled.
Poll
