This is pretty neat. We accomplished something similar but simpler. We use unattended reboots with randomised selection of time throughout 24 hours. Drain node with a script. Not perfect but simple to do and setup.