I have not worked in any company doing ab tests like that, it is always base vs variant. Option 1 vs option 2/3 and then if the feature wins, you can try option 2 vs option 3, if not they would scrap it, or blame it on the implementation and keep retrying until they either get false positive, or survivor bias makes it positive.

I am not defending it as I also believe that ab tests ruin the web, but if anyone does option2/3 at start they are not doing AB test at all.