We’re using Serverless where it fits and carrying on with our lives. I don’t feel the need to try and defend tools, because it’s exhausting. There are certain load graphs for which regular servers can’t follow the line in time, some for which regular servers can follow the line but would result in overprovisioning, and some in which request isolation is just a great thing to have. For these loads we use Serverless, for the rest we use normal EC2 or ECS.

Just understand the job and use the right tool for it. This fighting over Serverless and servers is like arguing over whether hammers or screwdrivers are better tools.