Happy thanksgiving! Leaving a Q before I forget - non necessary to answer today!
Have y'all thought about "inverse relations"? We've found ourselves needing to implement the inverse of a relation (prime example: orgs have repos, repos are owned by orgs). I think persisting those relationships isn't a good idea because it's denormalized information.
Have y'all stumbled upon something like this? Do you believe it's a "smell" of a problem somewhere else in the schema? Have you considered how to support this at the code level in a way we don't have to persist that duplicated relationship?