That's an interesting story re: early twitter. Are you certain a MySQL SLA-level support contract would have solved their scalability issues? My understanding is that it was high volume of highly followed participants that led to congestion. I.e. a design problem. Eventually solved by caching and fanning out replicated tweet data to all followers.

To see the issues around cloud and open source right now, you only have to look at who's winning: enterprise consultants such as Dell EMC, HPe, Accenture, IBM, etc.

They are the ones pushing broad K8S adoption at the edge. And in the next generation it will probably come down to a new crop of ISVs to solve issues around data portability, vendor lock-in, security, pricing arbitrage, and training.