Description
Standalone worker updates executor's state prematurely, making the resource status in an inconsistent state until the executor process really died.
In our cluster, we found this situation may cause new submitted applications removed by Master for launching executor fail.