The problem is simple to leave all this in the main layer requires programming in c ++ and requires hardforks for updates, on the other hand creating validation accounts the processes of the main network hive is only responsible for deciding which are the validation accounts and these Accounts will already have their own memory and disk systems to carry a separate process which allows better scalability.
Keep in mind that this will not only work for whive but also other dapps since it will allow these dapps to carry their own concensus systems in a 2 layer network.