Would it be possible to use a web-of-trust system to automatically resolve forks that occur when a node is offline?
This would just be a way of formalizing and automating the current plan to have nodes ask around about what exactly happened when they weren't looking, I think.