Not sure what I was linking to here. Watches are good for thresholding a single OID value, for aggregate comparisons (e.g. the average of all instances combined) I have perl script 5min cronjobs making REST API calls, then sending a trap for threshold violations.
Ideally I would like to extend the existing snmpd agent to include a new MIB OID that has aggregate values and then have Watches monitor those, but I do not have time for that.