The trouble with robotic repair missions is that it could work pretty well if you knew 100% exactly what was wrong, and exactly what else might go wrong while applying the fix for whatever went wrong. The more potential unknowns, the harder it is. It's much harder to make a robot with enough general flexibility that it could probably handle diagnosing and repairing an unknown issue, or handle something going wrong while trying to carry out a planned repair sequence. Bolts jammed, too loose, too tight, tanks of stuff springing leaks, electrical short in some unknown place, component overheating for unknown reason, all sorts of things can go wrong that are tough to diagnose remotely or with a special-purpose robot.
And you’d need to have it designed to be serviceable once deployed. I’d imagine there would be a few assumptions going into the design process, “once deployed humans won’t ever need to get to that..”
