Some of our agents on application owners servers became Yellow and or Red status. When we asked them to restart their processes. This fixed the agent health but the prior unhealthy agent unhealthy not go away. The process does not exist to kill when looking for it in task manager. Is there something we can do to remove these without requiring host recycles (which I'm not sure would fix it) on heavily used production app servers.
Are you displaying offline agents in the agents overview? This would show older ones that were connected. A screenshot may help.
Also, there will not be a separate process for the agents if we are talking about an application agent like Java. The agents are injected into the runtime of the application JVM and so will be running with the application and have the same process ID. You cannot kill an application agent - the only way to remove it is to remove the instrumentation string from the app startup and restart the application.