Do I understand it correctly that DuckDB would run embedded on the machine running Postgres (i.e through the extension), and this limits query processing ability to whatever that machine can comfortably handle?

What are the deployment implications if one wants to integrate this in production? Surely they'd need a much larger Postgres machine at a minimum.

Is there concern re: "hot neighbour" problems if the DuckDB queries get too heavy? How is that sort of issue potentially handled? I understood from another query that DuckDB is ran in a separate process, so there is room to potentially throttle it