As far as analytics its all done with tracking cookies now, which is another issue that needs to be addressed, but Im not going to miss referer should it ever really go away.
That said, I can see why certain people might see it as problematic... if you're browsing http://www.anarchistsbombmakingforum.com and somebody posts a link to http://www.fbi.gov, then maybe you don't want the FBI knowing you were at anarchistsbombmakingforum.com. But, still, barring other privacy problems, the FBI don't know who you are when you visit their site, just that you came from anarchistsbombmakingforum.com.
I have a hard time getting worked out about this though... for one, if you're surfing anarchistsbombmakingforum.com, common sense would dictate that following a link to fbi.gov isn't such a good idea (and a forum that fosters discussions of anything controversial should probably munge links to go through an anonymizer anyway) AND the people for whom this really matter are the people who have a referer blocking plugin installed in their browser.
The problems are numerous. I think I went through 9 revisions of the "naive protocol" (including, at one point, ditching the referer header in favour of HTTPS). Subsequently I realised that my tracking protocol was broken anyhow.
My conclusion is that there's no reliable way to track users visiting multiple websites using the standard features of HTML/JS/HTTP in the face of malicious users or publishers.
You have to fall back on traffic analysis.
I developed a successor technology which works better in many respects (but not all). As it's the subject of a current patent application I can't really go into much detail.