Tried this on a staging copy first and it behaved exactly as described. Confident enough to roll it to production now.

This matches my experience exactly. The hard part is staying consistent once the initial novelty wears off.

Strong agree. Good defaults beat a wall of toggles every single time.

This is the answer I wish I had found six months ago. Would have saved me a weekend of trial and error.

Could you say more about how you measured the result? I want to try this but I am not sure what success would look like.

Thanks for writing this up. Bookmarking it — the kind of thing I will want to reference again in a month.