I know this could be changed by writing another plugin, but by default when using NRPE, you'd have a separate connection and separate request per service entry. In our case, that could be 5 different NRPE connections per host just to check the disk. That can add a whole lot of overhead in terms of extra active checks.
In any case, I think it is easier to have the check_disk plugin check all the partitions at once (which it does by default) in only one active check. This is especially true if you modify the partition table--you'd have to change the arguments to check_disk and/or add a new service to Nagios. When all disk checks are in one service, it's a lot simpler.
Currently, check_disk does not have the features to do it the way I would need it done (such as checking all partitions with certain thresholds and providing exceptions for certain partitions), but I've also submitted some bug fixes and feature additions to sourceforge in hopes that they will make it into the standard package (after they clean up my code, of course ).
It looks like Nagios 3.0 will support $LONGSERVICEOUTPUT$ though.