Well, MO theory is mostly just a descriptive picture, or an interpretation of the math, so to speak. So was VB theory originally, but it's evolved into a quantitative method as well. ("Modern" VB theory and what they call the "Tight-Binding Approximation" in Solid-State physics)
They're both good pictures. The orbital picture in general is a very accurate way of looking at things. (But it's not something which exists. When you're talking orbitals you're actually talking about a particular mathematical description of a thing, not the thing) The matter of which orbitals exist (given a theoretical framework that usees them) and whether they're bonding or antibonding etc is exact and a straightforward mathematical result from group theory.
Neither MO theory or VB theory says anything about diamagnetism AFAIK though. Many textbooks state that VB theory cannot account for the paramagnetism of O2; this is actually not true. As Shaik explains*, VB theory doesn't really have a problem with this and never really did. (It seems it's all due to a statement Lennard-Jones made http://www.chemteam.info/Chem-History/Lennard-Jones-1929/Lennard-Jones-1929.html" , concerning Heitler-London VB theory)
There does exist compounds and situations where neither of these pictures give a good idea of what's going on though. Be2 for instance, which is more strongly bound than either MO or VB theory would lead one to believe.
In any case, these models are essentially the "language" of chemical bonding. Even if you're doing high-level quantum-chemical calculations with no direct relationship to either theory, the results still get interpreted and viewed through the picture provided by these models.
* Shaik, Hiberty, "A chemist's guide to valence bond theory", Chapter 1
