Now I have more maturity (and I have remembered this website exists) so I will give it a shot and try to start some kind of discussion here!
At the company I work we are using a full serverless environment (and using the Serverless framework you linked in your response) so we do not have to worry much about infrastructure. Our team is very very concise thus all all the technology demands end up on us.
Our goal by using Lambda Functions is to have less things to worry about while still dealing with low cost (the company is going to have its first client in a couple of weeks), having high availability and scaling on demand whenever needed without wories.
As the development team sees the business has got a solid basis and will eventually hit a high usage. Thinking about it and knowing all the boilerplate that comes with creating complete APIs via functions we discovered that there are wrappers that make it possible to create APIs like we usually do in a docker environment, for example (using Express for Node and Gin or Echo for Go, for example). By doing it we make it very easy to migrate to a docker/kubernetes environment in the future.
Another great point in doing this approach is that we stay in touch with technologies we are already used to and it is much much much easier to create APIs this way because of all the middlewares everyone needs nowadays.
Lambdas were initially created to carry very tiny and concise, objective workloads, but they are capable of handling a lot more work than that. I'll also add that even AWS is thinking that way because they have developed an open-source library that makes it possible to port Golang APIs written in frameworks to a Lambda environment (github.com/awslabs/aws-lambda-go-api-proxy).