Currently when Zeus receives any error reading from the command socket while waiting for a process to boot, it logs the error and then tries to parse the message anyway. Since the message shouldn't be valid when there's an error, we should instead enter an SCrashed state immediately. We could be throwing away useful debugging information bufferred from an earlier read but I don't think that'll actually happen in the existing Zeus code.
Currently when Zeus receives any error reading from the command socket while waiting for a process to boot, it logs the error and then tries to parse the message anyway. Since the message shouldn't be valid when there's an error, we should instead enter an
SCrashed
state immediately. We could be throwing away useful debugging information bufferred from an earlier read but I don't think that'll actually happen in the existing Zeus code.