The Class Leak Rule results shown on the Automated Analysis page is completely wrong. Attaching details and screenshots of the rule results for wldf.jfr analysis for better reference.
Classes Loaded - entriesLoad :: 5062
Classes Unloaded - entriesUnload :: 68
 
Class names and respective diff count i.e the number of times class is loaded minus the number of times class is unloaded:
java.lang.Object (258)
java.lang.String (95)
java.lang.Class (81)
java.lang.Integer (52)
oracle.jrockit.jfr.VMJFR (46)
 
Above are the 5 classes which are loaded maximum times without being unloaded. As per this information the rule should be defined as RED with the above information.
 
But the results are showing GREEN (OK) status and there are 5 classes mentioned in results which have diff count as (-1)
com.oracle.medrec.model.User (-1)
com.oracle.medrec.service.PatientService (-1)
com.oracle.medrec.service.PhysicianService (-1)
com.oracle.medrec.facade.PatientFacade (-1) com.oracle.medrec.common.messaging.MessageClient (-1)
This is existing code results and our test cases are also depicting the same.
 
Also, the rule is taking too much time to evaluate the results where the number of classes are in numbers i.e. in thousands. Sometimes its around 90 min.