It feels like, to me, that most of the comments in this thread that don't "get" serverless have never had to deal with things like maintaining the spaghetti mess of CM(Chef/Puppet/Ansible) that invariably results from trying to manage a traditional infrastructure deployment of any size.

Patching, dependencies, configuration drift, security audits/compliance... these are all non-trivial issues requiring non-trivial operations and engineering hours to solve.

It's fine if Lambda is too complex an abstraction for hobby code or a node.js todo-list app, but that doesn't mean it doesn't solve the right problems in the right problem spaces.