Currently in NPM, the selection of interface monitoring and the hardware sensors for those interfaces are completely independent of each other. When adding a device, and NOT choosing certain interfaces, you still end up with hardware health data on those interfaces (usually which is moot). Furthermore, even if the interface is present and administratively shutdown on the device itself, there's no dynamic identification that the OIDs for those particular hardware sensor are moot and should be skipped.
The only way to work around it currently is to add all hardware health (including those "in alert") and then manually walkthrough and disable every sensor that belongs to a shutdown or unwanted interface.
