Very interesting! I find especially such insights very important:

    • Erlang semantics say match clauses sequentially

                    BUT

    • Don’t have to if you’re smart!

    • Can group patterns and save testing

Note the "Don't have to if you're smart"! I see the lack of this insight frequently in the Prolog world, where implementors tend to argue against important features on the grounds that it "is" too inefficient. Well, the important question would be: How can we make it efficient? Examples of this are a compact string representation (which Erlang has, using raw sequences of bytes that can be reasoned about as if they were lists!) and a convenient interface for attributed variables.