The "succeeds just enough to survive" framing is the most honest description of feature debt I've read. Failed experiments are easy to cut. It's the mediocre ones with just enough adoption that become politically impossible to remove. If you can't name the specific user and the specific moment this feature is better than the alternative, it's not ready to ship.