DuckDB works well if * you have a small datasets (total, not just what a single user is scanning) * no real-time updates, just a static dataset that you can a

DuckDB works well if * you have a small datasets (total, not just what a single user is scanning) * no real-time updates, just a static dataset that you can analyze at leisure * only few users and only one doing any writes * several seconds is an OK response time, get's worse if you have to load your scanned segment into DuckDB node. * generally read-only workloads So yeah, not convinced we lost a decade.