Why is i20 = (Va-Vb)/20k, not Va/20k?
When an OpAmp is in equilibrium (ie: not in saturation) and in a negative feedback configuration, the + and - are practically the same. If we neglect the bias current for + and - it is fair to assume that vb = v+, is also = to v-.
So if the R20K sees va at one end and vb at the other end it is fair to mention that V(R20k) = va-vb,
since i20=V(R20k)/20K then the following is true: i20 = (va-vb)/20K.
It cannot be va/20k because v- is not zero, rather it is equal to vb
(See first postulate in phrase 1).
Of course, it would be true only if vb = zero