This is basically the same problems as in JENKINS-72200 for the disk space monitors.

When an agent connects but the JVM version monitor is configured in such a way that it should take the agent offline (e.g. paranoid++) the agent will be online 

Steps to reproduce

1. start Jenkins with -Dhudson.node_monitors.AbstractNodeMonitorDescriptor=3 (so that the monitors run every 3 minutes)
2. configure JVM version monitor to paranoid++
3. create an agent that is connected by running a command on it
4. run the agent.jar with a java version that is different from the controller
5. Visit the agents page in Jenkins (do not navigate there via the .../computer url), but e.g. directly click from the Jenkins start page)

Expected Result

The agent is taken offline

Actual Result

The agent is online and stays online even when the monitor in the background detects the version mismatch
When one now visits .../computer, the agent is taken offline immediately. Another way to trigger that an agent is taken offline/online is to visit the api endpoint of an agent.

The bug is basically that the logic to take the agent offline is implemented in the data method. This method is only called when the ../computer url is accessed.

The behaviour of the remoting version check is correct, the agent is taken offline directly after it connects without going to .../computer