Where PyIceberg works without Spark, such as serverless functions appending small batches, where it does not, and what to measure before scaling it up.
The real-time lakehouse is not one engine. It is a contract between streams, table commits, query paths, and freshness expectations.
REST Catalog V2 LoadTable work matters because clients and catalogs need explicit contracts, not optimistic assumptions.
The Rust versus C++ discussion is really about table-layer execution safety, interoperability, and performance envelopes.
