> Let's say a fresh fork happens causing a chain split. Both chains have 50% market cap and hashrate. How do you measure "popularity" at this point so that your prediction of "the most popular fork wins" is actually helpful? ETH forking away from ETC was a very controversial topic at the time.

You have to go forward into the future to evaluate the claim. The claim is that if you look at things... let's say 5 years later, one of those is going to be dead. If you find that one of the forks still has, for example, 20% as much hash power as the other, you have disproved the claim. So it's not a tautology.

Another way to formulate that part of the claim would be to say something like "the half-life of a fork is only X amount of time, so it's a very bad idea to try to pull one off".

> There are still Bitcoin Qt 0.8.x nodes on the network which are ~10 years old.

There's no way they're fully compatible. At the very least, consider that block that generated 184 billion bitcoins and had to immediately get the code changed. And look at all of these: https://blog.bitmex.com/bitcoins-consensus-forks/