I really appreciate your interesting and detailed take on this topic. I think the answer here is that the advertiser needs to do the research themselves, because they are frankly not entitled to this information by tracking my computer.
The example I like to give is a print newspaper or magazine. The New York Times is able to sell its ad space based on circulation -- and some weeks things might be slow and there might be tens of thousands of copies left unsold on newsstands, and other weeks there might be a terrorist attack where every copy of the paper sells out.
I think websites should do the same, sell local ad space based on average traffic over a window of time. If your site blows up into some megasite -- renegotiate your ad rates the next time they come up.