Basic Alert Manager shouldn't be used anymore. Open it up and remove any rules that you still have in there. And then test it again.
If Server B goes down at the same time as Server A, then the alert for Server B would be supressed. But if Server B stays up and Solarwinds can still get to it, then it may report on it separately. I've not tried this. All my group dependencies are in cases where if the parent fails, the child nodes are unreachable.
