In all previous firmware versions since 20100126 there was a bug which could cause random inputs to get stuck and not responding which in turn would require power cycling. E.g. in my case it manifested itself when electromagnetic noise from switching on/off electronic devices coincided with a specific value of an internal counter.
I'm a bit puzzled that nobody reported it, and I myself discovered it only a couple of weeks ago. Took me some time to track the cause.