We don't have features that take 10,000 lines of code. Our biggest features get split among a couple smaller PRs to review them along the way. I know there's a

We don't have features that take 10,000 lines of code. Our biggest features get split among a couple smaller PRs to review them along the way. I know there's a case for larger feature branches that accumulate an entire large feature before being merged into master, but in a world of microservices, it's just not a use case we've had to consider. And reverting PRs happens occasionally, but because changes are much smaller than the scale of what I assume you deal with, we never have many (if any) merge conflicts to deal with. We primarily deploy new features with feature flags, so it's vastly easier to just turn them off until we can fix the issue.