I'm a big fan of the Erlang platform, and I even kind of begrudgingly love the language, but Erlang didn't even have first-class immutable maps until 2015, and while I actually think the semi-prolog pattern-matching syntax is elegant and beautiful once you get a handle on it, I actually really dislike the whole comma-semicolon-period thing. (Personally, nowadays when I say I'm writing "Erlang", I'm actually writing LFE).
Also, the best webserver for Erlang is Cowboy (at least in my opinion), and the initial commit for that wasn't until 2011, and it wasn't really an awesome server until around ~2014. I'm not overly familiar with the Python ecosystem, but if I were to guess, the server frameworks probably were more mature than a lot of the Erlang ones.
I don't fully agree that it's difficult to benchmark Erlang, since it has a lot of awesome diagnostic tools available out of the box and moreover it even has stuff like caching and whatnot built-in with ETS, so I think that argument comes down to "we didn't really know what we're doing", but, you know, that's a valid enough reason to not use a platform if your job is to get something done. I wouldn't recommend most people write their new apps in Idris or Coq either, for no other reason than it is difficult to fully pick up and use those platforms easily.