Sounds like an excellent way to check to see if people are presenting alternate content like expert-exchange always used to.
If so, that's a tough problem. Distrusting your 'users' (content providers) and double-checking them is likely to generate a fair bit of hostility, especially for a non-opt-in service.
To be fair to Google, they do get inundated with spammers, so a certain level of distrust is understandable. The problem is that like Pierre mentioned this is very easily gotten around, which makes it fairly ineffective, and to anyone who actually uses traffic logs for any kind of stats tracking, any fake referrers can be quite annoying.
If they don't generate too much extra traffic, then I don't see any problem with spot-checking people. If it's a small amount of hits, then the faked referrer should just appear as noise.
Now, if it's a lot of traffic, then you might have a problem, but the post showed something like 2 hits. 2 hits from a crawler does not a spammer make.
No it's not. The IP addresses are Google's and thus easily targeted on the server for cloaking.
Normal Googlebot IP addresses resolve to names like crawl-blah.googlebot.com. These ones don't. You could use whois to detect them, but it's a lot slower than DNS. I guess that Google expects the SEO guys not to be smart enough to use something like pwhois to quickly check that the IP's on a Google AS (likely).
My argument still stands: the hurdle for anyone serious about cloaking is easy to work around.
There are some google ip addresses. Who knows how many google but registered in another company's name IP addresses they're doing this sort of thing from
That was my first thought too, and just because you notice this one doesn't mean they don't have more stealthy crawlers making sure you're playing fair.
How do they know that it was actually a bot, and not an actual person inside Google surfing from within IE6? The logs show two hits about a minute and a half apart, which seems much more human-scale than bot-scale.