I tried to use the Windows Service, which can be installed via a JLNP connection, to let the slave nodes automatically connect to the master. The installation seems to have a problem with shutting down the initial connection to master. So after a system restart of the slave it is no longer able to re-connect to the master because the slave is still marked as connected. It looks like that when the initial JLNP connection was killed the slave has not properly disconnected itself. Further restarts do not show this behavior. As of now you are forced to restart Jenkins on the master machine to let the client connect again.

Steps:
1. Setup Jenkins and a slave node via Java Web Start
2. Connect a Windows slave
3. Select 'File | Install as a service'
4. Wait for the service to be installed
5. Restart the slave

Once the slave has been restarted, the service will try to re-connect to the master, but the connection is not allowed:

Nov 12, 2014 11:11:05 PM hudson.TcpSlaveAgentListener$ConnectionHandler run
INFO: Accepted connection #618 from /192.168.123.3:60703
Nov 12, 2014 11:11:05 PM jenkins.slaves.JnlpSlaveHandshake error
WARNING: TCP slave agent connection handler #618 with /192.168.123.3:60703 is aborted: dummy-windows is already connected to this master. Rejecting this connection.
Nov 12, 2014 11:11:05 PM jenkins.slaves.JnlpSlaveHandshake error
WARNING: TCP slave agent connection handler #618 with /192.168.123.3:60703 is aborted: Unrecognized name: dummy-windows

I see hundreds of those messages in a really quick sequence.