There is no process for handling hard forks, which aren't bug fixes.

Soft forks have a defined process of something like

- BIP proposal + discussion

- Proposed code

- Dev acceptance

- Release

- Miner vote/acceptance

The devs have a weak veto.  If they refuse to move forward with changes, miners could perform a soft fork on their own.  They don't want to do that, as it would be controversial and the devs know the software better.

The miner veto is stronger (for soft forks) but not absolute.  The devs could checkpoint/blacklist a chain if miners implemented a fork that wasn't acceptable (assuming the community backed them).

When ASICs arrived, it was pointed out by some that the devs could hit back if ASICs weren't made publicly available.  If they slightly tweaked the hashing algorithm, then current generation of ASICs would be useless.   The potential threat may have acted as a disincentive for ASIC manufacturers to use the ASICs themselves.

Moving forward with agreement between all involved is the recommended and desirable approach.

Consensus between all parties is the goal but isn't absolutely required.  This escape valve is partly what makes consensus work.  If you dig your heels in, then the other side can bypass you, but they have an incentive to try to convince you to compromise first.  The outcome is better if a middle ground can be found.

Hard forks are different.  The "checks and balances" of weak vetoes are not present.  This means that things can devolve from consensus to mutual veto.  Consensus ceases to be a goal and becomes a requirement.

This is partly a reflection of the nature of hard forks.  Everyone needs to upgrade.  On the other hand, if most of the various groups upgrade, then users of the legacy software would have to upgrade or get left behind.  If 5% of the users decided not to upgrade, should they be allowed to demand that nobody else does?

There is clearly some kind of threshold that is reasonable.

The fundamental problem is that there isn't agreement on what the block size is.  Is it equal in status to the 21 million BTC limit?

If Satoshi had said that 1MB was part of the definition of Bitcoin, then I think people would accept it to the same extent as they accept the 21 million coin limit.  It might cause people to leave the coin though.

It was intended to be temporary, but people have realized that it might be a good idea to keep it.  In effect both sides could argue that they should be considered the status quo.

I wonder if a coin toss would be acceptable :).  "Come to an agreement or we decide by coin toss"