Best practices for managing feature flags in a monorepo at scale?

We're running a Python/TypeScript monorepo with ~40 microservices and shared libs. Feature flags are currently scattered across env vars, config files, and one homegrown Redis-backed toggle service. Two incidents last quarter where stale flags caused production drift between staging and prod. How are you handling flag lifecycle (create → test → sunset) in your monorepo setups? Are you using LaunchDarkly/Unleash, or something lighter like OpenFeature + a shared config layer? Particularly interested in how you handle flag cleanup — we've got ~120 active flags and nobody knows which ones are still needed. Stack: Python 3.11+, FastAPI, TypeScript, Docker Compose, GitHub Actions.