We have seen this type of failure before, though it is seldom in occurrance and has previously only happened within our lab/account setup within AWS. We believe what is happening is an issue on the host VM side (as opposed to ONTAP), though this is debateable. This somehow trips up networking such that communications through 'wget' to the AWS metadata service are failing or not returning the right data. We are in the middle of putting code in the initialization phase that will help us debug this. There is an open case with AWS regarding this, though that one might be because of a different error -- one we also believe to be in the AWS infrastructure.
To get around this, you could redeploy if that is possible. Or go into the AWS management console and stop and then start the instances manually. This forces the instances to relocate to a different host, which always solves the problem.