Track JNI failures and set a flag in the agent; check it at the end of the tests. This will ensure that any failure in the JNI subsystem is detected by checking for errors after each JNI call.