I believe that having self-selected teams results in the best results at hackathons -- at the expense of having a bi-modal distribution of results (i.e. the best teams do better, and the worst teams do worse). In general, considering the externalities involved in hackathons, I expect that most organizers will prefer to have the bi-modal distribution than one in which everyone produces equally mediocre results.

With regards to your other points -- they seem reasonable, but your post would be more useful to hackathon organizers if you instead focused on how to run a hackathon (i.e. what things they should do to keep participants and volunteers happy -- like having lots of food and/or clear expectations about BYOB).