It seems like the singular benefit is that DuckDB (or similar OLAP models) can quickly handle lots of expensive read queries on large datasets.

It’s not a replacement for a traditional RDBMS. I’ve never used it so I don’t know if it’s worth the effort to maintain instead of just using a Postgres read-only instance to run analytics queries but somehow I doubt it.

My guess would be that it has a few very specific use cases where it can provide some added benefit. So, I fully expect it to be crammed forcefully into software projects where it provides no tangible benefit for the foreseeable future. Just like cough MongoDB cough.