Idea #2 is particularly interesting. I think it won't work as is because it's exhausting to be on a treadmill of constantly refollowing people. Some people may welcome the opportunity to revisit decisions and shop around, but I suspect it might be going too far for many people. (I'm happy to be persuaded otherwise; chime in in responses if you feel strongly one way or another.)

The problem you're trying to address is that people rarely cull their followers. Every new medium gets users excited, but they end up following too much over time, swamping themselves, getting jaded, and leaving. I think about this every now and then; I even wrote about it 4 years ago: http://akkartik.name/blog/2009-05-19-21-30-46-soc

The solution I came up with was to measure time in votes. I built a site that looked like reddit but was a feedreader under the hood. Users need know nothing about RSS or XML, they would just upvote and downvote the stories they saw. As they upvoted stories they were subscribed to the corresponding sites, but subscription wasn't a binary activity, you could be 'partially subscribed' to a site, and the extent affected how many stories from it you saw on your frontpage.

When I built this, I saw something interesting. I observed many sites go through a common life cycle with their users. There'd be an initial honeymoon period where most articles were upvoted, and the site was close to 100% 'subscribed'. Then interest plateaued for a while at some lower level, then they lost interest, and a few downvotes brought it down to 0% subscribed, and they never came back.

It might be worth reconstructing this life cycle here somehow. Just an idea to throw out there. To reiterate, I agree with the problem. I'd like to make hubski a site that doesn't fail from too much success, that can handle lots of people, or people subscribing to lots of people.