It’s great to see a community member show deeper interest in CO.DELTA, I’ve cross-posted your question to this thread to keep the other one on topic (I hope that’s okay). NNS neurons are stored by the NNS Governance canister - you can read more about neurons here. Canister-controlled neurons are no different in this respect 
You can also read more details in our github repo if there’s still any confusion → CO.DELTA: Decentralised, community-driven due diligence for the Internet Computer
I’d be happy to elaborate further. Just let me know.
Thank you for your concern. Rest assured that changes to the CO.DELTA canister are run locally by team members before deploying. Note that the CO.DELTA canister has a very simple job, yet we do have utilities for quickly bootstrapping a local test environment. What test scenario are you concerned about Jefri?