Currently, Spark doesn't check hive.vectorized.execution.reduce.enabled. So there is no better way to disable it.