I've not had much experience with clustering, but have used NIS alot.
It seems your basically making a NIS master server, but the server is clustered to avoid NIS outage?
If thats the case, why you use cluster at all? Why not just have NIS slaves dotted around your network which sync from the master and you have better coverage?
If you have to use the clustered approach, I'm assuming the NIS service is down on one side, and when one node falls over, this triggers the NIS restart script on the other node? Have you just tried an /etc/init.d/nis start as part of the cluster handover in place of a bespoke script?