Techniques for physical machine management in distributed computing systems are disclosed herein. In one embodiment, a method includes accessing data representing an operating status indicating whether individual physical machines are operating satisfactorily in accordance with one or more target operating conditions. The method can then include based on the accessed data, identifying a list of the physical machines that are not operating satisfactorily in accordance with at least one of the target operating conditions but have not failed yet and determining whether the individually physical machines in the list are hosting one or more virtual machines. If one of the physical machines in the list is not hosting one or more virtual machines, overwriting at least one of an operating system, a user application, or a device driver on the one of the physical machines with a new copy.