Run a second instance of the API, watch its subscribers miss the events of the first, and share the events through Valkey, so every instance hears every change, even after Valkey restarts.
