I think the problem might be a TCP port conflict. The SUPPORT.SAR has all the log files, which is quite useful (and overwhelming!). In the dev_server0 we see just that it failed, as you reported. std_server0.out gives a little more detail, in that it reveals the manager that failed, which was ClusterManager. Still don't know why it failed, though.
Then, in the server0 defaultTrace_00.0.trc we see the error "Cannot open Join Port: [50020]" repeated a number of times, leading (I think) to java exceptions about being unable to establish loopback connections. Further down, a causing error seems to be "Address family not supported by protocol family: bind." I believe that is an artifact of the port conflict.
So, see if you can find out what else is trying to use port 50020.