I think you are right. This is a lot like manufacturing. It would seem to me, Replication by machines only becomes significant if the process is complicated by mistakes. Which may be the one thing missing in this scenario. The machines could go on replicating all they want, dominate the earth, and so on, until, the environment changes, or until the predator changes. That is when the machine's replication process would need some mistakes in order to survive. So it would seem that this process needs to incoprtate some unknown environmental variables and see how the machine can deal with those situations over several populations of machines. It would be interesting to see if the machines created errors to deal with the situations or if the erros that already existed were selected for.
