Personally, I reboot almost every time I made a config change. You never wanna be in the place where when troubleshooting something you hand-wrote an extra iptables rule, but never remembered to put it in your firewall script. Or hand-started a daemon, but forgot to add it to your init scripts. 8 months later you'll actually have to reboot because of some real issue, and suddenly you'll have to re-solve all your old problems. Yes this is a sloppy mistake, but as long as you suffer from the same 23 pairs of chromosomes as the rest of us you are likely to make it no matter how hardcore you think you are.
Besides, there's really only three types of systems.
1.) simple home/office user: uptime doesnt matter, just reboot.
2.) production server: load balanced or configured with failover, one node can be down for 2 minutes to reboot without a problem.
3.) rediculously high-end machines: all aspects of a change are made in test environments, then moved over.
Linux/Unix's ability to make more changes than say windows without requiring a reboot has no real business value. Any server who's uptime is too important for a 5 minute reboot, belongs in at least a failover-cluster.
bart