.... you always know where a servo is: at the position you tell it to go.
If it's not there, then there's a malfunction of the motor, or it's not strong enough. Either way that's a design error, and should never happen. Therefore having a pot to read back the position of a servo is unnecessary.
That's very simplistic, and neglects the fact that it takes time for a commanded movement to happen, even when things are all nominal. The next step in a process might rely on this process completing, and only closed loop control (which is what this is) will tell you that.
And then, when things do go wrong, which is not necessarily a design or manufacturing flaw, but perhaps unusual untoward and unforeseen circumstances preventing completion of a commanded movement, it's imperative that the system as a whole, knows that. It would need to prevent that next step happening, or perhaps reverse the incomplete movement back to a checkpoint.
If what you say is correct, there would be no need for closed loop control and everything would work on open loop, dead-reckoning.