I am working on network infrastructure, we mostly use command line flags to enable/disable features. The flag values are saved in the code repository and built

I am working on network infrastructure, we mostly use command line flags to enable/disable features. The flag values are saved in the code repository and built into each release. There is an internal initiative recently asking for dynamic config based feature control. The dynamic config is an RPC-based data push to the running services, which are expected to handle the config changes without restarting. It sounds less disruptive but may require careful cleanup for some features, in which case the cmd flag with a restart is much easier to implement.