That use case should work very well for a wide variety of hardware (and in fact I use it myself for a 20-node cluster). Issues may arise from networking and BMC unreliability, which the MAAS teams are currently working hard on addressing for this cycle.