I would like to be able to publish a hosted feature or table service directly from an Apache Parquet file item in ArcGIS Online. This would replicate similar functionality for other supported file types like CSV, zipped Shapefiles or File Geodatabases, etc. This would be especially efficient for tabular data and integrations with other data services. It would also reduce I/O, storage usage, and ensure data type preservation compared to something like CSV.
ArcGIS Online Data Pipelines already support creating or overwriting a hosted service from a Parquet file.
Ideally this would give Parquet files the same "publish" methods in something like the REST API or API for Python.