Ah, drat. I have an application that uses DuckDB to output parquet files. That application is by necessity disconnected from any sense of a data lake. But, I would love to have a good way of then pushing them up to S3 and integrating into a data lake. I’ve been looking into Iceberg and I’ve had the thought, “this is great but I hate the idea of what all these little metadata files will do to latency.”