Hi Pieter,
I believe Gavin plans to write a blog post about the hard fork process, but I'd like to debate this with you now, if only to give him material to work with :)
Your points look to me like the hard/soft fork debate in different clothes.
For example, we all agree that the rules of Bitcoin can be changed, and have been before (e.g. P2SH), with software upgrades.
When such a fork happens, any user who does not upgrade their node isn't fully verifying the block chain anymore. Their software might think it is, but it's running NOPs that don't mean NOP to other nodes. So there is a divergence in the consensus, it's merely been done in such a way that the node won't stop and print "hard fork detected" to the logs. It'll happily accept a block that violates the new rules, then wait to be corrected by miners.
So with any fork, hard or soft, there is risk to those who don't upgrade. They may accept a block, or even two blocks, that they believe are valid according to their old rule set, but which other miners would reject. The effect on double spending is much the same.
Now let's talk philosophy.