The Reports->CPU Usage report may be very slow of even fail in the UI as it tries to get live metrics from the agents. If now one agent is not reachable, the UI shows the yellow "I waited too long, but the operation may still go on" box and a popup shows "Failure". Clicking on "reload" does not help, as this also does not show the values last retrieved, but again tries to go out to the agents. The get live values should perhaps try to ping the agent and if this is not available within a very short time, abort and just return the last retrieved metrical value. Or totally switch to getting the last metrical value and then polling cpu usage and memory usage much more often than what we do today.