I haven't seen this problem before.
Can you please file a bug on this and include information about how to
reproduce the problem including the exact commands that were used to
create the domain, cluster, instances, and for starting the instances?
Please use the "admin" component for the issue.
Thanks.
Tom