Note that I understand what "Amazon S3" is; I'm asking only about the software that handles the HTTP interface.
If I check the HTTP headers for github.com, for example, one of the response headers is:
server: AmazonS3
Does Amazon actually have a web server that it calls 'AmazonS3', or are they just modifying headers from Apache/nginx/etc?
Server http header means nothing. It is an arbitrary product version string, think User-Agent but for the server. As such, it is untrustworthy.
Even if this was Amazon S3 (its not, that is a dynamic web site) there are ways to change the Server header.
Regarding something closer to what you describe, http Host does not imply a single backend server either, not that Amazon or GitHub would ever have a single server for anything important.
In general, I do not consider big name-brand tech companies to be good examples of off the shelf technology. An organization with huge traffic, a large engineering staff, and an engineering blog is more likely to write parts of their stack from scratch.
