Quoted from benheck:Since I have nothing better to do with my life, here's a graphic of how a node board could get bricked.
This is assuming the host CPU cannot electrically reset the node boards over the bus and must rely on commands telling the nodes to reset themselves.[quoted image]
If the main code is corrupted on flashing as theorized, what Stern should do is stay in the Bootloader waiting for a fresh programming command and not try and vector to the main code. This is pretty easily accomplished by first doing a CRC check on the main code before executing.
This would allow the end customer, us, to still be able to try and update the node board main code flash without anything being bricked.
Of course, this is all speculative as to what the real root cause is with the node board failures.