Worked on the githubnext/discovery-agent__apache__flink repository to enhance the stability and correctness of table creation processes, focusing on scenarios involving METADATA and computed columns. Addressed a critical bug where CREATE TABLE AS operations failed when METADATA columns were present, ensuring that virtual metadata and computed columns no longer overwrite existing source columns. Refactored schema creation logic to properly distinguish between physical and metadata columns, reducing the risk of edge-case failures in Flink table planning. Utilized Java, SQL, and schema management skills to improve reliability for downstream data pipelines, with all efforts centered on maintaining data integrity and operational correctness.