nice article - it's surprisiong how often people forget about checking UID/GIDs.
I'm not clear what benefit you get from using NFS, as against streaming the backup directly across the network. From the tar man page :
tar cvfb - 20 files | rsh host dd of=/dev/rmt/0 obs=20b
You can also accomplish a similar result using dump. I prefer 'sucking' the data, rather than 'squirting' it as you do with tar above, because you can run the backup set entirely from the backup server without scheduling jobs on multiple machines.
Backup script pseudocode :
FOR each machine to backup -
IF (ping machine to backup) THEN
suck data with a compression filter