yeah it is very likely dying on you.
and most likely it is a board or two.
not the controller.
definitely possible that the board is dying but for now it seems to be working from the local side, so i was wondering why the pool side hr hadn't moved. but it's slowly coming up now (>10.5TH/s) so the number on the workers>>workers page of kano.is must be a little bit longer average than what i expected.
Of course that can be high or low on the pool, but it takes a while to converge.
Also, of course, if you were almost continually mining, i.e. the stop+start was short, the 1hr hash rate will take longer to average out since it's not 'starting from now at zero' but instead averaging out the longer running lower hash rate with the higher new hash rate.