Polars seems nice but in my experience using it, the "lazy" APIs would still immediately materialize a ton of stuff in memory and had very spotty support on what data formats and storage integrations were possible with scan_* functions (though that was half a year ago and the support is slowly improving). It's frustrating, I mean really frustrating, to think I could solve a lot of my "scan through heinous amounts of data without any memory hungry things like window aggregations without blowing out my memory" with Polars and then watch my scan_thisorthat() call result in instant memory usage ballooning.

DuckDB on the other hand is wonderful and truly doesn't use any more memory than it really needs to.

Did you ever report these issues to polars developers?

I reported many bugs to both pandas and polars over the years and both teams tend to address relatively big ones. (Still have some outstanding pandas bugs that I think are a big deal but the devs disagree.)