our very own @surister explored methods how to efficiently load Parquet files, comparing different frameworks like pandas or Polars, then elaborating about PyArrow and multi-threading to parallelize converging from Parquet row groups into egress batches optimally, and finally rewrite all Python into Rust.