Question regarding re-installation of node and keys


I have an environment that runs mostly on ec2. Most nodes gets the default
(instance name) node name, but some instances (like db servers) have a
persistent node name, mainly for keeping values like ebs_volume if the
instance is terminated. AFAIK, the correct way to do it is to first create
a client, save it’s private key and copy the credentials to the server each
time I have to re-install it.
Since this is a highly dynamic environment I thought of making shortcut in
the form of deleting the client but keeping the node, and then letting the
reinstalled server register itself again. At first it seemed to be working,
but then I started to get errors at the end of the chef run (after all the
recipes completed):

[Wed, 16 Nov 2011 12:32:12 +0000] INFO: Processing
service[nagios-nrpe-server] action restart (nagios::client line 40)
[Wed, 16 Nov 2011 12:32:13 +0000] INFO: service[nagios-nrpe-server]
[Wed, 16 Nov 2011 12:32:13 +0000] ERROR: Running exception handlers
[Wed, 16 Nov 2011 12:32:14 +0000] FATAL: Saving node information to
[Wed, 16 Nov 2011 12:32:14 +0000] ERROR: Exception handlers complete
[Wed, 16 Nov 2011 12:32:14 +0000] FATAL: Stacktrace dumped to
[Wed, 16 Nov 2011 12:32:14 +0000] FATAL: Net::HTTPServerException: 403

I was wondering is there a right way to do it without having to register a
node before installing it?

Thanks in advance