first off, this cookbook kicks a**. SO glad you put the time into doing this.
I ran into an interesting situation where I started an new node, using recipe[riak::autoconf], and it failed because when autoconf picked a random node, it picked one where the node was not online...but other instances in the cluster were.
I could see this being an issue if during recovery i try to bring up another instance under a different node_name (perhaps riak3.west.ec2 rather than riak2.east.ec2), at which point it could then fail and I would have to ssh in and manually add it in. Not the end of the world but it is an extra step.
so what do you think of making random_cluster_member a bit smarter, and doing a ping and if it is up returning it, and if not, going to a different node?
Closing this out as we dropped the autoconf recipe. If you'd like us to bring it back or reimplement, please open a new issue, and I'd be happy to give it another look.