Lightning Talk: Feature Flagging at Scale: Problems with Flag Cleanup
Lightningat OpenFeature Summit Europe 2025 · on Tuesday, 2025-04-01· at 16:50· Level 3 | ICC Capital Suite 2-4·
summary
At large scale flags diverge from code: flags in the platform but not in repos, flags in code but not on the platform, and ambiguous Heisenberg flags that may be unreachable or mis‑targeted. Practical fixes include IDE integrations, ESLint checks, CI/CD gating, event exposure logs and static rewrites with Piranha (and its commercial offshoot Guitar). Real numbers make the pain concrete: Google runs ~100,000 experiments a year and Uber reports about 800 PRs/month from automated cleanup. Testing explodes combinatorially.
AI-written summary · about
chapters (7)
- 0:00Feature flagging at scale
- 1:43Problems with stale flags
- 4:51Solutions for unused flags
- 6:38Heisenberg flags cleanup
- 7:41Testing flags in production
- 9:24Automated testing challenges
- 10:27Conclusion
related talks
30mA Tale of Two Feature Flags
08mThe Art of Feature Flagging
56mWhen Feature flags go bad
06mFeature Flag Foul-ups
53mDr Sheldon Cooper Presents: Fun with Flags
44mFlags of Flexibility - Unlocking Rapid Iteration
12mRevolutionizing Software Development
11mFeature flag use cases you haven't heard about yet