There's a pretty nice set of aggregate check statistical functions - the basics of min/avg/max/last are covered, but one that would be helpful is median.
The value of median aggregate is that it tends to be resistant to bizarre/false readings. Say you're reporting the average drive temperature from a collection of 36 drives and a few have faulty sensors that always report "0" or -65,535 or + 65,535, that can really thrown the existing aggregate functions way off - while median wouldn't change much and would be a much more accurate number to base AC, airflow, etc. decisions on.
There is also statistical value in having having a grpstdev function. For example evaluating grpstdev for ping time would have statistical significance for subnet/branch issues that grpmax wouldn't (single messed up device) and grpavg/grpmed might also not show as clearly. I see this as less clearly helpful as grpmed (proposed group median aggregate function) but still useful if someone is already messing around with the aggregate function code.