In this case what you monitor is typical behavior. Given a normal bandwith of number of process (which needs to be tuned for the system), you may want to be informed or warned if there are suddenly an x percentage more than usual and maybe get a critical message if the number of processes exceeds yet another treshold.
It may indicate a problem that perhaps can be corrected before the system runs out of CPU or typically memory (or bump into another limit like a kernel parameter that prevents new processes from getting started). Another thing is that if the system runs out of CPU or memory, this metric will provide you with additional information that may help you determine the nature of a problem...