At face value this may seem the case, but a closer reading reveals the appropriate limitations that prevent this paradox. As described (p 83) - fast-casting is defined as "a response to an attack or other surprising event" (emphasis added). It's not a prescient interruption such that the attack/event never happens, it's a reaction once the attack/event is perceived and is already on its way in.
About halfway through that section, the language begins to assume that this response is purely defensive - to counter the attack itself (as opposed to, for instance, just smoke the other mage before the attack ever happens). So, the fast-cast spell by Mage B can be used to stop the incoming spell from mage A, but it appears that it is not intended to be used as a coup-de-main to prevent that spell from ever being cast.
Why? For exactly the reasons that the OP found this disturbing and paradoxical, that Mage B lost initiative and so should not be able to attack Mage A first, "fast cast" or not. If Mage B makes his fast-cast roll, he aborts his normal attack and defends against the incoming spell that he perceives, but he cannot create a spell that stops the casting of that spell before he can see it to react to it - he lost initiative, it's the "attack or other surprising event" he's reacting to, not the threat of that attack.
And this makes sense of the timing and situation: Mage B loses the initiative, but is not completely helpless against the attack (as is the case in some RPG's, where the victim simply stands there cleaning their fingernails until the attack is resolved.) An opportunity for such a response not an easy or guaranteed, but it's possible (and more possible for some magi than others).
(The obvious counter-argument is that "the casting" is the surprising event, that Mage B can, after losing initiative, address the threat of the casting of spell and doesn't have to wait to react to the spell itself. To take this to its logical extreme, then an enemy coming around a corner is "a surprising event", so Mage B should be able to fast-cast a spell before the enemy even comes around the corner to respond to that, maybe to create a wall to prevent them from coming around the corner ... and that's just silly. A mage responds to the attack itself, but not the threat or promise or potential source of that attack. Nothing else makes sense. 
I suppose that if Mage B wanted to ignore the incoming spell, they could cast an attack spell (or anything else) - but that has its own risks, and at that point there's no reason to fast-cast unless Mage B doesn't think they'll be around after Mage A's spell hits.)