@ Flying Binghi: you're talking about the past. I (and Backpacker) are talking about the future and its rapid rate of technological change.
Call me an optimist if you will (how dreadfully unfashionable!), but Backpacker's suggestion of technical solutions seems entirely feasible. Computer vision systems are advancing at a tremendous pace, and are already far more effective at detecting visual anomalies (albeit with a modest false-positive rate) than humans. Of all the computer vision problems that are being worked on (facial recognition, human behaviour recognition, etc.), spotting an aircraft that's on a collision course is relatively straightforward algorithmically. It's quite plausible that within three years, we'll have UAVs that are better at detecting potential VFR traffic conflicts than pilots. Combine that with TCAS, and UAVs programmed to take deconfliction action with some knowledge of the other traffic's likely behaviour (e.g. observing the Rules of the Air), and I don't see how they'll be a significant increase on the mid-air collision hazard as it is right now.
Pragmatically, it may well be easier for the UAV guys to use this technology to allow UAV operations outside CAS, than to try to force massive restrictions on airspace that's currently class G. The technology will be extremely valuable operationally anyway, so it's probably going to happen, and class G will probably remain much as it is.
An interesting potential spin-off of this is computer vision systems for piloted aircraft to assist lookout. Cue flame war about "what's wrong with the Mk.1 eyeball", "lulls pilots into false sense of security", "keeps eyes inside the cockpit", etc...
