It seems the results (knowledge silos, review cliques, and rubber stamping) would be inaccurate often since the heuristics can’t actually capture what they say they do? I would be concerned about users not having much faith in it once it returns one too many false signals.
Not saying it’s common, but those situations do exist.