I figure it'll basically be two kinds of work that can do that shit, for a long time:

1. Minimally tethered, highly mobile robots that can climb all over the place and do industrial welding, in high places, underwater, on the side of rockets, in construction work, etc. These will be ruinously expensive, for a long time. But they will be worth it, because you can literally have the next shift come in and put the control-suit on, without even having the robot waste time climbing down. And you never have to worry about human safety protocols.

2. Robots that are more bound to a single "station" area. Kind of like the robotic welding machines that do autonomous tasks in the automobile industry, but set up for full human control, from the next room. These would probably be much cheaper, since they wouldn't have to move on their own. You'd have to set the work in front of them and pick it up when the task was done. But, again, they'd be cheaper than the mobile ones.

Either way, this would be something that would be confined to relatively specific use cases, for quite a while.