If we compile statistics, we can look for the points where nobody has ever meaningfully recovered.
We do. This is where the guidance to stop resuscitation after 15 minutes without a rhythm comes in. Unless you're a child who drowned in cold fresh water, of course, or an adult who apparently died of hypothermia. The problem is that there are so very many different sets of facts, and people are far more resilient than you can imagine. And in the heat of the moment, we tend to opt to fight rather than to let go. Which is actually OK, I think.
For your example, ECMO can only be useful for acute lung failures including injury. It's useless in chronic cases where the lungs simply aren't going to improve.
Well, yes, ECMO is probably a bad example, because it's by definition an acute therapy that can't be continued more than a few days, at least at the current state of the art. Even there it's a bit questionable in the case of chronic disease exacerbated by acute cardiopulmonary collapse from a (presumably) reversible cause. But other therapies, like the simple $10K/day ICU bed, are much harder to argue against, unless you've given specific instructions. It reminds me a little of the old instructions for tuning a carburetor - turn the screw until the engine dies, then back up half a turn. Most of the really futile ICU cases I've seen didn't START as futile cases, but they sure ended that way.
Most other western countries have a bit less tendency to heroic medicine than the U.S.
I'm not actually all that impressed with medicine in "most other Western countries" as a touchstone for our own. Every country has its own social norms and conventions, all of which fold over into health care. We tend to value privacy, autonomy, personal space, personal choice, and hope for recovery more than most, and it costs us a lot of money.