It's not the Death of PaaS, it's just getting put in the right place
I'm certain that we aren't seeing the "death" of PaaS, but are seeing a reality develop in the market around it's relative import and the value of different approaches to supporting application owners. "Pure" PaaS solutions are certainly valuable under the right conditions, but come with all the same implementation and learning curve challenges inherent in any new and complex technology. More importantly, they often require rethinking the applications themselves, which is something that I find hard to swallow as an engineer. When you step back and look at the evolution of IT in a more general way, one stark reality stands out... the long tail. Enterprise developers need to find ways to take advantage of the enormous value possible with orchestration and abstraction, but in ways that bridge the gap between the future and the current realities.
In my opinion, Amazon, Google and Microsoft have it right... they offer the power of modular services (think RDS, DynamoDB, etc...), coupled with composition *options* (call this the "PaaS" if you want to) for the application owners is the path forward.