Higher curation, cheaper downvotes, and even mild/moderate superlinear (superlinear I don't support) are all part of changing the algorithm. They all shift economic incentives and the incentives are very much part of the algorithm. Without input from external actors (votes) the algorithm does nothing and those external actors act based on the incentives. You can't separate one from the other.
Most obviously cheaper downvotes are a big change because they dramatically change curation incentives toward voting for content that has a lower risk of being downvoted. This means the curator must consider factors predictive of downvotes and/or net votes (including valuation and perceived value) and not just factors affecting upvotes (visibility, misterdelegation vote targets, paid votes, etc.)
You've made a pretty strong argument for an independent downvoting system, something I already thought was a good idea but a low priority; you've convinced me it's more important than I was giving it credit for.
But nothing you've argued has made me think that its benefits won't accrue just as well under 75/25 as they would under 50/50.
Fair enough. I agreed that outside of 100/0 we can't really make any strong absolute claims about what are the right numbers. We can make relative claims about shifting incentives in a direction (which is the reason for most of the support behind 75/25 -> 50/50), but that's not quite the same thing.
So given that and the concerns about "omnibus bill" forking, why not try independent flagging first and see how it works?
I already said it was fine with me. Given that the track record release cycle for hard forks is 1.5 years+ I'm a little concerned about moving too slowly but I don't strongly oppose it.