It appears that instance03 stopped responding to GMS heartbeat and was reported as suspect and then failed.
This can happen when the GC is running for long periods of times and not allowing for an instance to respond to its
heartbeat within 6 seconds. However, since the instance was also confirmed failed, that means that GMS was not able
to create a new TCP connection to the instance and ping it with 10 seconds of the missed heartbeats.
It would assist in evaluation if the DAS server log is attached since that provides an overview of unexpected
failures. Then the server log and gc log of that instance that failed first should be inspected to
see if it is long GC and/or OutOfMemory issues that is causing that instance to seem failed. Or if there is a fatal
exception showing that instance has crashed.
NoClassDefFound error reported in http://java.net/jira/browse/GLASSFISH-15428 seems more likely to occur on a
failing JVM due to OutOfMemory issues.