What I support most strongly is that we do substantially-delayed hard forks with conservative values. Then doing hard forks regularly isn't such a big issue.

For example, Bitcoin Core can be immediately modified to increase the max block size to 2 MB on a specific date 2 years from now. I think that pretty much everyone would be basically OK with this max block size (and even higher values might be widely acceptable). By the time the change actually takes effect in 2 years, everyone will already have upgraded because very few people use 2-year-old software. Businesses and users won't have to go out of their way to choose one fork over another, and so there will be less room for messiness.

Then if a really nice academic study convincingly arguing that 5 MB blocks are safe is published 1 week after the 2-year-delayed change is added, another 2-year-delayed change can be added right away with very little extra cost. After ~2 years, the max block size will increase to 2 MB, and then a week later change to 5 MB.

Yes, 2 years is a long time. But I'm confident that Bitcoin will survive that long with 1 MB blocks.