Service endpoints are cached excessively

We are running Openstack with a cold standby service host and an ip-alias that is moved during takeover. This leads to failures after takeover since the service endpoints are cached.

To repeat:
- Setup:
- Two hosts running openstack (host1 and host2).
- One service ip-number (serviceip) assigned to host1 at the begining.
- Steps:
- Configure the plugin to use an endpoint on serviceip
- Test the connection using the GUI
- Preform fail over so seviceip is located on host2
- Test the connection again using the GUI
Now you get the error:
Connection not validated, plugin might not operate correctly: Failed to connect to host1/10.2.0.130:9696
java.net.ConnectException: Connection refused (Connection refused)

Workaround to get things working: restart Jenkins

Our failover happened a week ago

Unfortunately I didn't copy the full stacktrace before restarting Jenkins and can't easily recreate it since openstack is in production and making test failovers needing company wide planing.