To test a distributed system, there's not a lot of value to simulate all the many different conditions that can happen on a machine. From the system's perspecti

To test a distributed system, there's not a lot of value to simulate all the many different conditions that can happen on a machine. From the system's perspective, you don't really care what happens on a single computer. The conditions you need to simulate are (a) the machine being abruptly gone or (b) the machine still accepting requests, but being very slow in returning them and maybe (c) machine returning incorrect results. Seen from the outside, everything that can happen is usually (a) or (b), and with ECC memory and reasonable software (?) hopefully never (c).