Hey John. Right on! In the 1980s Tandem Computers (I worked there) was already building "non-stop" computers that ran two processors (separate chips) in lockstep and checked that they produced the same output, and even ran triads of CPUs that voted. This technique detected what I guess are now being called CEEs and prevented them from propagating into and corrupting the data. Meanwhile, another custered node would pick up the workload so the system worked non-stop. (CEO Jimmy Treybig, a Texan, liked to demostrate this by shooting bullet holes into a running system.) These systems were used extensively in banking a stock exchanges. May still be, as far as I know, though Tandem was acquired by HP in 1997. Anyway, here's agreeing with you that it seems like those techniques can be incorporated within mulit-core systems to mitigate the CEE problem.