Consider new framing for chain splits

[moneyball]

AJ Towns feedback: "There's two ways you get a chain split: deliberately or accidentally: it's deliberate if you move forward with a consensus change despite a enough people objecting to it that two independent chains can survive; it's accidental, if everyone wants the same thing (whether that be the pre-existing rules or some new rules), but through bugs or lack of foresight, you get nodes following/mining on top of different tips anyway. The "bounty" case is only relevant when there is a deliberate chain split -- the bounty funds can only be claimed via a chain following legacy rules.

For the deliberate case, I'd say the risk of chainsplit is highest when you've got medium adoption/agreement: when adoption is very high, you might avoid a chainsplit because the people who object give up, and when it's very low, you might be the one that gives up; but with lots of active support for both sides, that's when a BTC/BCH-style split is most likely.

For the accidental case, when the bug is "too few people have upgraded", then if few miners have upgraded and also few node operators have upgraded, you've got a huge problem, though one that can be fixed by having people upgrade and manually reorg to before the problem. If few miners have upgraded but many nodes have, the result is just a high orphan rate, possibly including long orphan chains; with miners who haven't upgraded becoming much less profitable, and users getting slower confirmations, but a surge of orphan blocks doesn't really create a new coin. If most miners have upgraded but few users have, things should be fine, as long as more users upgrade over time."