People often talk about scaling in a lopsided manner. Meaning scaling is only up, well scaling works both directions. For an architecture or runtime to be scalable, it needs to shrink as the demands placed on it fall. Scalability works in both directions.
Yes and no. Realistically you often have other constraints. E.g. for a lot of code I wrote at my last^2 job there would be no value in scaling below a single micro EC2 instance (I ran each service on its own instance and got value out of doing so), and almost all languages can scale down that far.
