That pack only surfaces what’s available through the Coda API. By raw data I mean I can work directly with the JSON/codabuf files and read e.g. how much space exactly each row and column takes (to quickly identify doc bloat issues without having to nag the Coda support), read system tables and columns, identify data leaks in published forms, read the true dependency graph to identify performance bottlenecks (although, I’ve seen it’s there but I haven’t yet examined it) and so on. Many ideas from the many use cases I’ve experienced with clients who were pushing Coda to the edge.