There's a great podcast/interview with the creator of duckdb. He's pretty clear of thinking of the use case as more or less equivalent to mysql but for aggregated queries. I think trying to use it in place of something like a fully fledged postgres server might get weird, less because of any issues with duckdb and more because that isn't what it's designed for.

Podcast/interview link for anyone interested: https://www.dataengineeringpodcast.com/duckdb-in-process-ola...