Worked on the apache/flink repository to enhance resilience and observability by introducing a dedicated CoordinatorNotExistException in the Java runtime. This targeted feature improved error handling and exception management by explicitly modeling scenarios where an operator’s coordinator is missing. The implementation involved refactoring runtime logic to throw and process the new exception, while updating CollectResultFetcher to treat these cases as expected conditions, logging them as debug messages instead of warnings. This approach reduced log noise and clarified failure semantics, resulting in more maintainable and predictable operational behavior. The work demonstrated focused use of Apache Flink, Java, and robust exception management practices.