We're discussing whether to stop logging the "Stable cache miss", as it turns out we missed a major use case which made the message much noiser than expected. When the stable revision and the latest revision have the same wikitext content and none of the templates have changed (for example, a vandalism change is rolled back and both revisions are still pending), then FlaggedRevs will take advantage of the main ParserCache and does not leave an additional copy of the data in the stable-version parser cache.

The data collected so far is interesting, however. It shows how often this use case happens—and my guess is that it also confirms that there are no major leaks where we're not taking advantage of the stable cache when we should. Here are the numbers: