One flag, no deploy: switching diagnostics and providers at runtime

Do you force your own failovers in production, and if you do, did it fire the way you expected the first time?

I run a small app on Cloudflare Workers on my own , so when something breaks it's just me, and a deploy just to turn on logging takes too long while a user is waiting. So I put my debug logging behind one Workers KV key with three states, off, admin and all.

I gave a talk at Feature Flags and the recording shows the forced failover live I did, and the challenges I encountered while doing it and lessons learned. What is your view?