SSH Build Agents: controller restart leaks agent processes. I noticed that we have two process-sets running on all our build-agents:

jenkins  19108 19099  0 Jan19 ?        03:42:22 java -jar remoting.jar -workDir /var/lib/jenkins -jar-cache /var/lib/jenkins/remoting/jarCache
jenkins  24558 24548  0 Feb17 ?        00:29:57 java -jar remoting.jar -workDir /var/lib/jenkins -jar-cache /var/lib/jenkins/remoting/jarCache
 

I did some troubleshooting based on the date and time when the new process was started on each agent, and it equals the date and time when we installed a new plugin and restarted the jenkins-controller.

Preliminary search in the plugin code did not show any effort of trying to detect/kill any earlier existing agent-processes.