why not just go full bore to duckdb?

because method chaining in Polars is much more composable and ergonomic than SQL once the pipeline gets complex which makes it superior in an exploratory "data wrangling" environment. While DuckDB now has its own new expressions pipeline implementation it's way worse than Polars'. DuckDB has other advantages though but Polars is a much cleaner Pandas replacement. Earlier versions of DuckDB were also crashy whereas polars feels carved out of granite.