If the developer does not need to deal with a long-running process abstraction, and is billed only by actual compute time and not idle time, I think the consensus is to call that "serverless"

Thanks, finally a somewhat clear definition. It's still a misnomer though.

True. Point to remember that client-side != serverless.

S3 isn't a server, it's an object store. This uses Lambda (the serverless part) to fetch source data on demand, combine, reproject to web mercator, and send to the client.

There are varied definitions of serverless; it isn't "as serverless" as the browser loading data directly from S3, but an AWS marketing person would call this usage of Lambda serverless.