In the example you gave, the source Table is backed by Kafka, and the sink Table is backed by Iceberg. Flink will process the data being read from and written to these Tables without mirroring it elsewhere.
However, Flink may need to create a materialized view in order to execute your query logic (e.g., if you're doing a GROUP BY aggregation). If so, that data/state will be stored in the state backend configured for your Flink cluster, e.g., RocksDB. That state is not exposed as a Table -- you can't access it directly.