+1 for “Programming Erlang” and “Designing for scalability with Erlang/OTP”.
Those two books opened my mind to a lot of ideas and were very influential for me recently.
Don’t fear the Erlang side, the concepts (of course) map very well to Elixir and you’ll get a very solid (or a better) understanding of OTP behaviours.

For the distributed side and scalability, I guess the main takeaway is “it depends”. On your scale, requirements, tradeoffs, team… The book is very nuanced.

A little personal example : after reading “Designing for scalability with Erlang/OTP” I changed my SLA to 99.5% for my SAAS and added two “maintenance windows” in my contracts that allow me to perform planned downtime. That looks like an outdated practice and very low availability today.
In practice systems stroll around 5 nines and I don’t use those windows. It’s easy at my scale. But availability has first to do with human expectations and human availability. If there’s a global DNS or worse outage while I’m driving someone to the hospital, there’ll be downtime. Without people on call or 24/7 shifts I can’t promise more, whatever tech and distribution strategies I’d be using.

That’s the kind of very immediate practical takeaway I got from those books .

Elixir in Action has a chapter on distribution, and Engineering Elixir Applications has a chapter on deploying distributed applications. The Little Elixir and OTP Guidebook also has a couple chapters on the subject.

Distribution is a VM feature first and foremost. You‘ll be using the exact same primitives across the whole ecosystem no matter the language used. Only higher level libraries will then start to make a difference.