I assume you are suggesting making a network call to the backend to invoke the logic. I wanted something that could respond quickly (as checkboxes are clicked and unclicked, in a long form), so I didn’t keep “call the elixir backend” as an option for this case.
You never mentioned where the infrastructure will be located, if you plan on making a local application where you embed elixir and elm on same machine, then using some of your options might be possible (even though they don’t make much sense, there will be no visible performance gain).
If you need an active communication channel between server and client, then just use websockets. There is the phoenix_channels library that abstracts a messaging interface and luckily there is a client in elm also (if that doesn’t work well you can always use the official JS client), however at this point I would just use liveview as it already does what you want to implement from scratch.
Let’s say Amazon us east data center for backend, and California for client browser. Want library function calls (or network calls) to complete within 500 ms. I guess a websocket could support that need. We already use phoenix channels through elm on some pages.