I would be happy to trial the 50/50 as you know, especially after playing a little on Smoke.io and watching large accounts post much less and spread more. I still believe at this stage of the platform's life cycle, experimentation will reap much larger future gains than the possible short-term losses through errors made.
If there is a decent framework set up that allows for short trials (a couple months) with systematic monitoring and adjustment proposals (revert if absolutely necessary), then the possibility to develop robust solutions is increased. It seems that everyone is looking for a solution to commit to that fixes it all, and there is far too much complexity and unknown to likely find that solution at this point.
Yes, I don't expect everything to magically work on the first try. But I think we can see some decent improvement even in the short term.