They seem so bad for factory work though? One of those robot arms would work so much better than some shitty humanoid robot.

If a factory can afford robots, they already have acquired it. Industrial robots excel at their work already due to them being extremely precise already. If you need transportation robots, there are already ones that euter run on embedded rails or are already fully self-driving using wheels. Humanoid robots solve no issues that the industry hasn't already solved. It would just be a robot that would be less stable compared to any other transportation robot nor as precise as stationed ones while also more complex, and thus easier to break down, with the only upside it being that it's more of a generalist, but that is also sort of a moot point because a human could do it still cheaper.

The real use case of humanoid robots is very niche, with it being in environment where classic robot models fails, that being an environment that cannot be modified for classic robot use (e.g. mountainous terrain) where flying is not a viable option. After all, the human body, and the bodies of quite a few animals, excel at climbing rough and steep terrain whereas most, if not all, currently commercially available robots fail at it, or at the very least do very poorly.