Larry,
I'm not familiar with Helga's writings, so I cannot comment on that specific case. In general, I do not see the problems you see. There will always be edit conflicts, and giving a small, however selected elite control to resolve them seems like an awful idea. Sure, many times those who are shouting against the majority are just cranks. Sometimes, however, they happen to be right.
Note that one of your favorite "bad examples", Everything2, is an example for a community that has been completely eroded by a supposedly benevolent elite (albeit not a random, changing one).
I am, like many others, a big believer in the concept of "soft security". If we want Wikipedia to develop in a certain fashion, we should try to enforce our rules through peer pressure. People who violate NPOV should be educated about its purpose. Antagonistic statements of the "if you don't like it, leave" sort do not cool down conflicts, they drive and fuel them. Express respect for the other person's view, and try to find a way to integrate it without violating NPOV.
If this kind of behavior was more effectively trained and practiced by Wikipedia regulars, I believe we could deal with seemingly destructive newbies much more rationally. But the prevailing attitude by many contributors seems to be: "If the other child plays with my toys, I either take them and go home, or I find someone to complain to". If we want to be the adults on this playground, we should behave accordingly.
Note that Everything2 does have a few good ideas, and one of them are so- called mentors. Newbies are taught the ways of the site by old-timers. To make this work, however, we need an improved internal messaging system and a mentor selection process.
Article certification mechanisms we are currently discussing may serve as a further incentive for people to come around. If we get this right, the biggest honor a Wikipedia author can receive is to contribute to an article certified by a high number of users -- something that is worth striving for. Hopefully, this will motivate at least some people to examine their behavior.
IF and only if *all* else fails, I believe randomly chosen samples are a bad way to make final decisions. Slashdot uses such a scheme, and you probably have noticed how well it works. Decisions should not be made randomly but by those who care about the subject in question and have all the necessary information. A random sample tends to make uninformed decisions following a certain average pattern.
I'm not entirely against content-based bans, but I believe open voting would be necessary in such situations. See my previous posts on the subject for ideas how to implement this properly.
Regards,
Erik