Jar包以systemctl服务运行时无法正常使用Kinesis流，直接执行却正常

我碰到了一个棘手的问题：直接用命令java -jar filename.jar运行我的Jar包时，它能正常读取并消费AWS Kinesis流；但把它配置成systemctl服务在同一个环境下运行时，每次启动后都会毫无报错地触发优雅停机。

下面是服务运行时的关键日志内容：

• Tomcat started on port 8092 (http) with context path '/'
• Started Bootstrap in 3.906 seconds (process running for 4.903)
• Initial state: [StreamConfig(streamIdentifier=stream_name, initialPositionInStreamExtended=InitialPositionInStreamExtended(position=LATEST, timestamp=null), consumerArn=null)]
• With failover time 10000 ms and epsilon 25 ms, LeaseCoordinator will renew leases every 3308 ms, takeleases every 20050 ms, process maximum of 2147483647 leases and steal 1 lease(s) at a time.
• With failover time 10000 ms and epsilon 25 ms, LeaseCoordinator will renew leases every 3308 ms, takeleases every 20050 ms, process maximum of 2147483647 leases and steal 1 lease(s) at a time.
• Initializing LeaseCoordinator attempt 1
• Waiting up to 20 seconds for shutdown to complete.
• Starting lease cleanup thread.
• Number of pending leases to clean before the scan : 0
• Starting LeaseCoordinator
• Scheduling periodicShardSync
• Initialization complete. Starting worker loop.
• Worker shutdown requested.
• Elected leaders: fd1ceed5-3f45-42a3-aef2-...
• Worker 5d35a20a-35a1-46f4-8ae2-... has successfully stopped lease-tracking threads
• Stopping the lease cleanup thread.
• Shutting down leader decider on worker 5d35a20a-35a1-46f4-8ae2-...
• Successfully stopped leader election on the worker
• Shutting down periodic shard sync task scheduler on worker 5d35a20a-35a1-46f4-8ae2-...
• All record processors have been shutdown successfully.
• Starting worker's final shutdown.
• Shutting down CWPublication thread.
• CWPublication thread finished.
• Worker loop is complete. Exiting from worker.
• Completed, shutting down now.

我已经尝试过下面这些方案，但都没有解决问题：

• 修改Jar包及相关文件的权限、所有者，让它和直接执行时的用户保持一致
• 将故障转移时间调整到10分钟

备注：内容来源于stack exchange，提问作者Earl Cee