At the moment we have a fairly large install of PRTG (just over 1500 sensors) and we are currently in the process of reducing 'noise' from alarms - i.e. make it so an alarm is actually an issue to be dealt with rather than something that can be ignored.
One feature that we would like to request is some way to deal with flapping sensors/devices.
We have a lot of remote sites being monitored, some of which have very flakey Internet connections or flakey VPNs. We want to know if these go down for a significant period of time, but not if they drop for just a few seconds.
Now there is the facility to increase the polling interval but this leaves it very much down to chance whether you get notified or not - i.e. if the connection happens to be down for 10s at the point when your sensor checks it will alarm.
One solution would be have an option to select 'this sensors must be down for x seconds before it alarms'.
We can already do this with ping sensors to some degree by setting the number of pings to a high number (5 second timeout and 24 pings means the device has to be down for 2mins before it warns then 2 further minutes before it alarms), but there is no way to replicate this with things like SNMP traffic sensors on the same device.