Scaling cloud infrastructures can experience lag during the periods of high activity, where other assets have to either be added, or become active. This is the compromise for a cost-effective solution that scales. The company could go for a system that is absolutely overkill on assets at all times, in preparation for those brief peak moments. But this is expensive, and unlikely to be taken by most companies. Only case you would want to use one of these is if you have a sensitive or critical service that MUST remain online. Stock exchange servers, military servers, bank servers, etc. come to mind for this criteria.