Are humanoid robots the answer for a lot of use cases?

The resson industrial robots are the shapes they are is because it fits the job. A stamping machine foes not need a face or legs. It feels like humanoid designs are taking that process in reverse- come up with the design as a general model, then hammer it into a use case.

Even if the premise is the classic "robot butler/caretaker", will the best design necessarily be humanoid? Two legs and feet make self-stabilizing harder than 3 or 4 wheels, maybe a single main arm and a set of multiple clamps would be better than two symnetric hands for fidgety movements.

I also feel like the near future is fleets of narrowly scoped devices for resilience and better task fit. If the Roomba is broken, but the Kitchen Delivery Drone still works, it won't derail your dinner plans the way an all-in-one Roise the Robot Maid failing would. The Roomba can also stay 10cm tall to fit under the furniture even if the KDD needs a refrigerated compartment big enough to hold a turkey.

And let's get ahead of it: no pain emulators. We know how it turned out for the Kaylon.