Roughly 100 million page hits per month on the web app. Ran smooth as silk.
Do you mean that you are fetching data from an internal API and sending server-rendered HTML response to the browser? I always naively thought GitHub worked this way. Build a GraphQL/REST API that exposes all of your system's internals, query/mutate through it and send an HTML response in the "frontend" (Ruby on Rails etc.). Frontend, but in terms of an application server that in turn accesses everything through an internal API.
Most of the Javascript was not for interaction, but for serving ads, really.
It's not quite a backend and it's not quite a frontend, but it sort of sits in the middle and (hopefully) helps you achieve lower latency. I haven't used it myself but I think it's a promising idea and based on your post I gather a similar concept worked very well for you.
How did they compare with open-source ones (Drupal et al)?
Do you find serverless CMSs to be more useful than traditional ones or vice versa?