it's straightforward for a PR to get stuck in a review loop as the author issues "review ... again" over and over, cycling between the same set of reviewers. it'd be useful to store an association of reviewers to PRs in memory, with a configurable TTL, so that if a reviewer has been recently assigned, they can be excluded for the next retry.