I would say most of the tooling around Parquet sucks. A lot of simple use cases (head, sample, merge, etc) are locked behind rather heavyweight ecosystems (Hadoop, Spark, or other JVM dependencies). Newer tools are starting to develop that make the situation better.

Is Parquet better/faster/stronger than keeping everything in schemaless CSVs? 100%, but it has historically meant that I have to make trade-offs to benefit the tooling rather than how I want to interactively approach my data.

Most people do not have "big data" problems where the performance differences of Parquet vs csv matter.