“ If your x noise was greater than your y noise, you generally wouldn't bother taking the measurement in the first place.”

Why not? You could still do inference in this case.