I suggest that a global variable be added, which states how old a revision can be (when not still the top rev) and override the default. That way, outdated quality revs won't always override newer stable ones.
I don't think the quality-expiration setting would be a good thing to do. Its presence is not apparent to editors and pretty much unexpected when it takes effect. It doesn't have clean lines. It's hard to pick a value which is good for most articles in a wiki. Editors will think "sometimes they see old ones, sometimes they don't; who knows?".
On the larger question, a blanket policy of having Examined versions take precedence over Sighted versions for all articles seems inadvisable to me. The effect may well turn out to be to "disenfranchise" normal (autoconfirmed) editors. As soon as an article has a version rated as Examined, normal editors no longer have any more power than anonymous IP editors to affect the published article. The article could be said to have been "poisoned". After this point, only those with Reviewer rights can get a version published. This is subject to accidental or casual "driveby rating", or more abusive version-pushing by Reviewers.
I think a better approach is generally use "at least Sighted" as the policy for a wiki as a whole, and then to use per-page criteria to explicitly set higher requirements for some individual articles. The criteria can be set to the Examined/Quality level, or Featured, or whatever is appropriate for the particular article. If it later is decided the higher requirement is no longer warranted for an article, it can just be set lower, e.g. back to Sighted. (Contrast this with Quality-over-all, where it would require finding every Quality version and re-rating them to just Sighted, thus losing the information, just to fool the mechanism and unpoison the article.)
-RS