Actually, with my statistical hat on, that's a very fair point. Momentary spikes would be easy to detect - there are a whole suite of statistics aimed at identifying trends which are different from the norm. Add in standard deviation and median to your mean measure, and you've pretty much got a way of identifying where there have been peaks of interest (you would not be able to tell if it was abuse or some exciting new news for example).

I'm not the greatest modeller in the world - I tend to do it at a very basic level - but basic modelling may be all that is required here. I'll try and find the time to have a play with some numbers...