Is there any proof it never has in any situation on any site in the SE network? I agree the way I contrived this particular case is unlikely and may never have happened, but it is a possibility.
I went extreme as the quick simple example, but does it work better if the +25 are from users with 10+ years of professional experience with the topic on hand in multiple environments with multiple projects versus -15 from users with less than two years of experience and have worked for a single company on a single project? Much more possible situation to occur, but also far longer to type and read.
Nevertheless the point I was making still stands; without a statistically significant number of votes to offset “odd” situations like those proposed, the actual vote count does not provide actual relevant data to the end user as they cannot know the “value” of those votes. They are simply left with the potentially wrong conclusion that more votes is better or that posts without negative votes are better than those with negative votes (if they can check).