Comment by KraftyOne

To be clear, a DBOS worker isn't just one connection, but an entire application server that can handle many requests concurrently and batch/bulk write database operations. Past that, I don't really think we're disagreeing: there are clever optimizations (batching, offloading coordination where possible) you can do in both a library-based and workflow-server-based system to scale better, but at truly massive scale you'll bottleneck in Postgres.