Of couse, there are a few exceptions to this rule, like HNs favorite; Factorio, and Kerbal Space Program, Prison Architect, Rimworld are a few other examples. But most Early Access games released are in reality excuses for releasing something unfinished early, and only do small changes after that.
the mvp model works well for games like this that aren't really story based. I'm always happy to start a new map in factorio to see what got added in the latest version. I'm not gonna go back and replay a 40+ hour rpg just to see what it's like with a few less bugs or a couple new cutscenes.
Only the "maximum viable product", the best possible version of the game within budget and technical limitations, is relevant.
The fact that the devs improved it significantly afterward for years is unusual, especially since that game in particular does not rely on microtransactions which would benefit from a strong base game.
this is pretty much the only release I'm looking forward to this year, and I very much appreciate that they are pushing it back until they feel it is finished.