This is an issue similar to CDITCK-74. The following tests in 1.0.2.SP1 fails with GlassFish
org.jboss.jsr299.tck.tests.event.observer.enterprise.EnterpriseEventInheritenceTest.testNonStaticObserverMethodInherited
org.jboss.jsr299.tck.tests.event.observer.enterprise.EnterpriseEventInheritenceTest.testNonStaticObserverMethodIndirectlyInherited

Farmer.observeEggLaying [1] still tries to record using "this" and EventPayload.recordVisit() [2] still uses the passed object's class.
This may not work in environments where the EJB Bean is subclassed as is the case with Glassfish. For instance in GlassFish, we see the following
[class org.jboss.jsr299.tck.tests.event.observer.enterprise._LazyFarmer_Serializable, class org.jboss.jsr299.tck.tests.event.observer.enterprise._Farmer_Serializable]
Daemon Thread [http-thread-pool-8080(1)] (Suspended (breakpoint at line 45 in EnterpriseEventInheritenceTest))
EnterpriseEventInheritenceTest.testNonStaticObserverMethodInherited() line: 45

Need to fix the test so that is agnostic of EJB implementation strategy – either see if the recorded classes are assignable from the original EJBs or have the 2 EJBs have a method that returns their respective names.

[1] http://anonsvn.jboss.org/repos/weld/cdi-tck/tags/1.0.2.SP1/impl/src/main/java/org/jboss/jsr299/tck/tests/event/observer/enterprise/Farmer.java
and
[2] http://anonsvn.jboss.org/repos/weld/cdi-tck/tags/1.0.2.SP1/impl/src/main/java/org/jboss/jsr299/tck/tests/event/observer/enterprise/EventPayload.java