I agree: RuntimeError, which is a generic base class for application runtime-type errors.

ArgumentError is more for situations where the wrong sort of thing was given to a function, rather than a bad value, but a type not capturable in guard syntax, such that a FunctionClauseError is inappropriate.

In particular, changing the environment variables given when running your program would allow the same function to start working, so there clearly wasn’t a problem with the argument given to it: there was a problem with the runtime it was executing in.

Sources:

• Usage of ArgumentError in the Elixir language
• Usage of ArgumentError in Elixir libraries
• Usage of RuntimeError in the Elixir language
• Usage of RuntimeError in Elixir libraries
• Behaviour of guard failures:

  defmodule Foo, do: def bar(x) when is_list(x), do: :baz
  try do: Foo.bar(1), rescue: (e -> IO.puts(e.__struct__))
  # Elixir.FunctionClauseError
  

RuntimeError