But I do not believe that experts should have any special powers in the editing of articles.
Rather, I think they should be encouraged to act in a pure review capacity, assessing the existing work of Wikipedians, and making recommendations for improvement. This might also be partially implemented through flagged revs, and I could also envision a type of button at the top of articles that says "see last version assessed by an expert".
My point is actually that for majority of articles on science-ralated (and possibly some article on humanity-related, here I understand the situation less) there is nothing to review - they are either stubs or non-existent. Somebody needs to write them. You can consider this as a kind of review if you wish.
Cheers Yaroslav