At Fresha, Emiliano and his team tested whether Snowflake and dbt could become a direct write path into Iceberg: Amazon S3 storage, Lakekeeper as the REST catalog, and StarRocks as a downstream query engine. The tables were valid Iceberg, but the operational contract was weak: refresh semantics, catalog operations, file layout, table properties, compaction, and downstream performance all required workarounds or repair.

This talk is a production case study exploring the difference between format compatibility and system interoperability. “Supports Iceberg” is not enough, because the writer determines whether the table is actually portable, maintainable, and fast.