when this error occurs again. try disconnecting slave & re connect (jenkins->manage->node), if that fixes then it must be problem with the jobs you are running which are exhausting disk/memory and you need to find out fix for that.
explanation: when disk/memory runs out & slave is constantly trying to be in connection, few of the classes are loaded in jenkins server and few doesn't that is why you see, jenkins is up, and your job as well is running but the reporter failed to report because it needs that class.
how I can confirm this: I was going through same problem 10 days back and reading jenkins node's log gave me clues (there was an interruption exception and then node came back but not properly)
and if this is true for you then it is not junit plugin's issue, can be closed.