I like the 2 ending questions. Here are my thoughts on both:

1. While the Opt-in/Opt-out model is great on paper. Natural human greed will kick in, and most writers will at least eventually opt-in - proof is in the fact that most top writers happily let publications syndicate their articles and pay with money/traffic.

Even if some writers resist, with most opting on, this pressure will tip most in.

More so, how hard would it be for Medium to track opt-in/opt-outs and accordingly adjust earnings?

2. I think this becomes tricky territory because no search engine trains on a single piece of content or a single writers' pieces - the training, assimilation, and output is likely sourced from a myriad of places/people's content.

So I am not sure how this would work in reality