Somewhere in there, I argued that the "best" options for tanking were actually worse, as the game lost excitement and drama. What's an RPG without those things? It's a mechanical exercise. Imagine Robo-Rally without all those traps and all those bumps.
Well, what makes drama? What's the right level of drama vs competence? That lead me into a theory about institutionalized risks in an RPG, and that lead me into a technique to measure that risk.
End result? My little theory went and predicted the general consensus of what classes make the best tanks. Whoah.
Even better, this theory can also help you fix overpowered classes, design classes to proper power level, design monsters, assess overall party capability, and even explains why players and spectators act as they do.
The theory boils down to this: Any game requires institutionalized uncertainty. Without that uncertainty, you don't have a engaging game. As a natural relaction, in order to gain an advantage, players seek to reduce uncertainty as far as possible, but only in their own favor.