At the risk of a) being completely wrong (hey, it happens ); and b) butting in when most of the rest of you are already bored with this discussion, here's my two penn'orth:
The OP's question seems to be:
Magus A (eg initiative 7) casts a spell (eg Ball of Abyssal Fire) at Magus B (eg Initiative 3). Normally, B would fast-cast a defense, preventing the original BoAF from reaching him. Then he gets to cast a spell on A, who in turn can also try a fast-cast defense.
However, what happens if B fast-casts an offensive spell at A? And is it better to be first or not?
Being first is best, in all 3 versions that I can think of.
Version A
The troupe rules that any reasonable response can include an attack and not just a defense.
A's BoAF will then hit B (if it penetrates, etc) and hurt him. This will happen regardless of any of the fast-cast events.
B needed a fast-cast 'initiative' that beats A's original one (eg of 7); his offensive fast-cast spell may or may not penetrate/harm A.
A will get his own fast-cast defense spell, which in my opinion still needs to match his own original initiative. It probably isn't worth A's while to do a fast-cast attack, but he can if he wants to. This would allow B's fast-cast offense to take effect which will now happen regardless of any of subsequent fast-cast events.
If A does do a fast-cast offense, then B would get another fast-cast opportunity, now needing to beat a much higher initiative total (7+6=13). Again, defense or offense. A can counter again, ad infinitum, until someone fails to achieve the higher and higher initiative requirements.
However, each round where an offensive fast-cast spell is attempted, there are more spells that have not been countered and will take effect.
After all that, B (if he is still alive - this is why it is better to go first - and not too fatigued) gets to cast his own actual spell on initiative 3 (NB the rules actually state that failing to achieve the required speed prevents any further spell-casting, but I think it should only prevent further fast-casting. And, if the troupe agrees, might only apply to each triggering event). Which A can then defend against. See above.
If B does a 'normal' fast-cast defense, I would not allow A any fast-casting - that B defended himself is not a surprise. The surprise or unexpected event would be if B attacked A.
Version B
The troupe rules that a fast-cast defense can only be a defense and no attacks are possible. This is contrary to the opening paragraph on fast-casting on p83, but implied in the bottom paragraph of the middle column where it uses the word 'defense' when using fast-casting against magic (it is not clear if alternatives eg offensive magic are allowed or prohibited). This method has several advantages, not least being the time each combat round takes.
Version C
The troupe decides that casting spells is interruptable, like it is/was in D&D 1st ed-3rd ed (I've never played 4th). In which case that troupe needs to come up with rules about how long formulaic and spontaneous spells take to cast, how easy it is to physically interrupt them as well, and probably several other ramifications. These rules do not currently exist, as shown by Callan and C'hound in their discussions, above.
I would go with version B, myself. It's the simplest and stops everyone playing non-magi getting really bored during combat. C requires some house rules, but might end up being pretty good.
As an aside, I lived through several iterations of similar arguments about instantaneous spells in Rolemaster. The publishers eventually came up with a rule that a mage could only cast a max of 4 of these a round...which did not help much since spellcasters with that type of spell were 4 times better than those who did not have them on their spell lists. What worked was a rule that a mage could cast one instantaneous defense spell in addition to his normal action of one spell of any type, since most defense spells were instantaneous.