I think it's just that no one has documented it yet. I haven't yet tried py4web locally, much less hosted, but I'm confident that you'll find it straight forward.
My experience setting up web2py on AWS was pretty simple. It's much like the Centos/Redhat environment. Nginx and Postgres were no problem. I did start with the web2py's Rocket server, but the certificate support is easier using Nginx. My certificates were from Let's Encrypt.
If you're doing command line stuff, there are a couple places where AWS uses the Ubuntu-style commands, but installation is with yum. Service start/stop might have been Ubuntu-style, or maybe it was cron.
I had everything on one instance, and had no problem with hits about every second once I figured out the Postgres stuff to stop it's memory footprint from growing. I think it took creating an index to do the scrubbing, but it's been a couple of years since I touched that, I'm not a DB analyst, and the server went down with the ship when Startup Land finally claimed its victim.
I didn't do clustering/roundrobin-ing. About the only failures I had were due to the Postgres memory naive setup, and that went away as I climbed the learning curve. For mail, which was mainly for reports, I used the native sendmail rather than the AWS mail server. I didn't get into to Elastic shards because we did pretty simple traffic analysis (number of requests per country, using IP data, how often a client returned to the site, etc).
/dps