Flink high-availaility in standalone cluster, kill the jobmanager process. But the jobmanager can not recover later

Question

Flink Version 1.17.2

Cluster Detail Computer A： JobManager TaskManager Computer B: JobManager TaskManager

Zookeeper start by the bin/start-zookeeper.sh

I am sure that jobmanagerA and jobmanagerB is success.

I want to know why i kill the jobmanager processs in computer A, and then jobmanager in computer B handle the job.But, many time later, jobmanger process in computer A can not recovery from failure automatically.