Is it safe to say CLs that roll dependencies shouldn't be surfaced as culprits? For now I'm thinking we can implement some basic heuristic analysis on the culprit to check if it's a roll, then bail out of auto actions if it is as they're likely to be not the true root cause.

My first question is: is it really a false positive?

I'm basing my assumption on the CLs that instigate the rolls have historically always been reported as wrong by developers/sheriffs when it comes to flakiness

Can you follow up with sheriffs for this particular one?