This is something that was previously discussed on MSO for Stack Overflow:
The idea was posted very early on by someone who is now a high-ranking member of the community and actually a current moderator.
It has a score of -9, with a rebuttal answer of score +24.
Now, admittedly, your proposal is slightly different. You're suggesting that people receive this weighting for upvotes in a specific tag, and only in that tag, as opposed to a general upvote weighting based on reputation in general. However, many of the old arguments still apply:
People are still attracted to high numbers. This system would allow the categorized experts to bias the community vote, which is exactly what the SE design aims to avoid (voting is as fair and impartial as possible).
Upvotes are still not (necessarily) the same as expertise. Upvotes, even in a single tag, are partly a measure of quality but also a measure of time, and someone who is not a real expert but has consistently provided many "decent" answers could easily amass more upvotes than someone who is a real expert and has only provided a few answers (albeit amazing ones).
We'd still be telling new members that, essentially, their votes don't matter, and at our current stage new membership is extremely important.
If the weight of downvotes were increased, the idea takes on sinister overtones, allowing privileged users to game the system by burying other answers to questions they themselves have answered.
In addition, any proposed solution based on tags suffers from its own problems:
Tags aren't always reliable. Over time we've put a major dent in the awful [technique] tag, but haven't even gotten started on the equally awful [cooking-techniques] tag. A lot of people just don't use good tags, especially newbies, and right now our community isn't great about retags (that's OK - even Stack Overflow has its issues).
Tags can change, and can be changed by users with fairly low reputation for that matter. This is a good thing, it helps keep the site organized, but presents several difficult conundrums when you start introducing important privileges based on tags. Should people without "tag expertise" be allowed to retag, even if the tags were blatantly wrong? If "tag experts" have placed votes, and then that tag is removed, are the votes then recalculated? What if the retag happened months later and the damage has already been done (i.e. by biasing the other votes)?
I completely agree with the premise that we need to recognize subject-matter experts. I think one of the things our friend ignored is that we already do this with the interweaving of reputation and privileges; subject-matter experts have veto power on questions.
Perhaps that's not enough. Perhaps we need some form of veto power on answers as well. It's tricky, though, because we don't close questions merely because they're bad - they have to be a particular kind of bad, and most of those criteria don't apply to answers. So such a feature would have to be thought out rather carefully to make it meaningful and prevent abuse.
I'm open to discussing better ways to retain experts - and in fact I think the Stack Overflow team is already discussing this issue internally and at length - but unfortunately, I don't think this is a good way. It's just fraught with problems, several of which could undermine the very factors that make these sites so appealing in the first place.
Anyway. I am sure Darin has his reasons for departing. We don't know if it's temporary or permanent. My guess is that, as our site passed the "peak of inflated expectations" in the Hype cycle, new questions started to dwindle and he wasn't seeing that many new and interesting ones. This too shall pass, it's part of the life cycle of every new site or product. In any case, I'm quite certain that the community recognized the value of his contributions as evidenced by his average of 5 upvotes per answer.
I also want to mention that just because it appeared on somebody's blog does not mean that the "evaporative cooling effect" is actually a real or interesting phenomenon. There is turnover in every community, and many communities are occasionally graced with the likes of subject-matter celebrities just as many restaurants are occasionally graced with the likes of Hollywood celebrities; that does not necessarily make them a "hot spot" for such or prove that they are any more successful than other communities.
The biggest problem with that post is that the author, by defining success on his own terms, has rigged the debate. It's choice-supportive bias and we have it too. Insular communities that over-emphasize expert/celebrity status do have higher immunity to the brain drain, but they also have far less immunity to groupthink and general decay. I have been a member of a few communities now that were great with member retention, never really suffered from trolling or low-quality members, but eventually just died of "old age" - people got bored and that was that.
What's ironic is that no matter how open your system is, people will still complain that it over-rewards the elite. Capitalism, supposedly, rewards the big corporations and the super-rich. But if that were true, then why does it work any better than feudalism? It's because anybody can compete. They can start from nothing at all, and if they're good enough, work their way up to the top.
SE sites are, or at least try to be, like free markets. The rags-to-riches has happened on Stack Overflow a number of times. It happened to me; I had a mostly inactive account, one day I decided to start actively participating, and in about 6 months I was near the top of page 3. Then I got, well, bored, but while I was on my way up I found myself neck-and-neck with another guy who did keep at it and he's now in the top 20. That's what's great about these sites - that people who aren't necessarily recognized experts can become recognized.
Perhaps some feel that they aren't "recognized" enough, but it doesn't seem to have deterred a lot of those people from continuing to actively contribute. And even if we do want to be "better" at recognizing them, I don't think that giving them "super-votes" is the right way, because while that may make those experts feel better about sticking around, it ultimately hinders the ability of other members to grow.
Perhaps what we need is a form of in-game currency. We sort of have that with bounties. Somebody who's amassed a huge amount of reputation can, without really giving it a second thought, fire up a bounty and award up to 500 rep points for what they think is an amazing answer. The bounty award even shows up next to the post, and unlike votes, bounties are not anonymous - you can see which expert awarded it.
Personally, I think that's actually enough. The recent-ish improvements to the bounty system actually provide a really great and useful way for people who consider themselves experts (and whom the community has recognized as experts) to highlight other experts and help promote them to expert status.
But maybe we need a bounty-like system for votes as well, that might give experts the ability to bump answers up the tree - at some cost to them. Possibly worth investigating as well.
However, in my opinion, not really worth investigating yet. We're barely into the growth stage; what we especially need is to attract new experts and not necessarily worry extensively about retaining the ones we have (unless they are seriously dissatisfied, obviously - but I don't think that's the case at the moment).
