I have a couple of questions regarding the implications of the whole thing / any resources that I may have missed would be very welcome as well.
Can bots/crawlers be reliably identified and excluded?
Can a fairly unique user ID be generated in a reliable manner (e.g. with browser fingerprinting)?
Should it be implemented at http server level or with php?
a) If http, can info from the actual content be retrieved and sent over to google?
b) If php, would caching become an issue?