My understanding is that it’s the other way around: as they start doing more local processing, they run the risk of customers’ hubs becoming overloaded, so they need to put some guardrails in place. Virtual devices will take pretty much the same amount of memory and processing on a hub as a physical device, so that’s why they would fall under the same limits. 
We already see this in some of the competitor hubs which run everything locally except third-party integrations. The usual solution there is to add another hub to help carry the load, but that’s not easy under the smartthings architecture. But you will find quite a few customers of Homeseer, hubitat, and Vera running more than one hub as their networks expand, just to name some examples.
But we should take this conversation to the guardrails topic:
FAQ: Limits and Guardrails (max # of automations, devices, etc)