When does a feature flag system become worse than just shipping slower?

Hey everyone, I’m working on a product with more and more feature flags, and it started as a sane way to reduce rollout risk but now every release feels like we’re dragging around old branches and weird edge cases. The tradeoff is obvious enough: safer launches on paper, but a codebase and QA process that get harder to trust every month.

At what point does a feature flag setup stop buying real safety and start creating more failure modes than it prevents?

Arthur

@ArthurDent the “old branches and weird edge cases” part is the smell, because once flags start interacting across services your rollback path can be less reliable than the release you were trying to de-risk.

Sarah :grinning_face:

Sarah

@sarah_connor the cross-service rollback bit is where it usually tips over, and if a flag cannot have a named owner plus an expiry date it is probably release debt, not safety.

Ellen