Ideally, the rankings should be associated with a time period. Something ranked high... but all those high rankings are from 2 years ago, and there's been none since... that's a different signal that current high rankings and current code changes.
If a project has a total of 10 reviews, it's probably best to not just take the average of the 2 newest reviews. On the other hand, if a project gets dozens of reviews a month, taking of the last two months or so would totally make sense.