Actually this is a great feature. How it basically works is that any interface marked as unpluggable will collect information when it is up and not when it's down. More importantly when it's down it won't trigger an interface down alert. So for instance you wanted to monitor every interface on a router in a data center, but you were only concerned with collecting stats when those interfaces were operational and you didn't care about actually being notified when those interfaces were down. That's what unpluggable is for. In our case we are a networking group and as such want to track utilization and error performance stats on all the data center router interfaces. However, the group that manages the servers attached to the servers is constantly taking them down with out notifying us. If we didn't have unpluggable set on all those interfaces then every time the server group took a server down we would get a critical interface down alert. So unpluggable saves us the aggrevation and still allows us to provide usage stats to the server group. Hope that explains it a little better.
This feature was specifically designed for user access switches. Since folks take their laptops and such home at night, a number of ports on your user access switches will go down or red in the evening. Since the status changes to down, any node status = down alerts you have configured will fire. By marking user access switch ports as unplugged rather than down eliminates the status change to "down" and thus avoids the alerting.
I would strongly urge you not to use this for anything other than user access ports. The reasoning is that if you mark an uplink or a server to go "unplugged", you will miss an actual outage if an important interface goes down.
Not trying to be argumentative with the previous reply, but it is not recommended to use this on server ports.... specifically not production server ports.
Hope this helps