Some of our rendering processes take multiple days to complete, and the blackbox software we use doesn't have a pause button. So it's not that we're in need of

Some of our rendering processes take multiple days to complete, and the blackbox software we use doesn't have a pause button. So it's not that we're in need of 99.99999% uptime, but there's actually never a moment where rebooting a machine would be convenient (or indeed cost us money). Being distributed over nodes means I can reboot them and the processes are not disrupted.