Scaling from a few dozen sure is a big undertaking, but the vast majority of streamers are streaming to only a few dozen at a time. Something a relatively strong VPS can handle. Once your audience is above that, then is it not unreasonable to expect actual broadcasting infrastructure? I think the argument being made is that twitch as it is, with basically infinite free video encoding is massively unsustainable.

Platforms like Floatplane exist, which is sustainable. So its not an unfathomable problem to solve, its just more complicated than text.