The Rishonim offer various reasons why even though normally when two sets of witnesses contradict each other the result is a draw, when it comes to eidim zomimim we believe the second set of witnesses and discard the testimony of the zomimim. Ramban and others suggest that we are only left with a standoff in a case where Witness Group A says Reuvain stole $100 and Witness Group B says Reuvain did not steal $100, challenging the veracity of their testimony and. However, if Witness Group A says, “We were in the store at 2:00 on Tuesday when Reuvain stole $100,” and Witness Group B says, “You were with us on Tuesday and not in that store,” Witness Group B is not saying anything about the veracity of Witness Group A’s testimony regarding Reuvain. Since there is no direct contradiction, the Torah accepts their testimony as accurate.
The Derashos haRan gives another interesting reason. If Witness Group A wanted to lie, they could in theory pick any day of the week and any time of day and claim that is when they saw Reuvain steal. Witness Group B does not have the same luxury. They have to testify about the time and place set by Witness Group A. There is no wiggle room for them to concoct a story about that precise date/time in advance and hope no one can prove them wrong.
So the law of eidim zomimim makes perfect sense, right?
Yet in Bava Kama 71 the gemara says according to Rava that eid zomem is a chiddush, i.e. something unexpected, outside the norm. Given the explanations given by the Rishonim, what makes believing the second witnesses unexpected? Isn't it the logical thing to do? True, Abayei disagrees with Rava, but it seems far-fetched to say that Abayei and Rava were arguing about whether the reasons given in the Rishonim are compelling and make sense or not.
This question is asked by my wife’s grandfather, but in truth the Lechem Mishneh (Hil Eidus ch 18) beat him to it (as he notes himself). I think we can answer the question by clarifying what this whole concept of a chiddush is. My thinking on this was driven by the discussion in ch 3 of R' Moshe Avigdor Amiel's introduction to his Midos l'Cheiker HaHalacha. Here's his reduction ad absurdum argument: if you define chiddush as something that defies logic or common sense, then you've just about obliterated the term because every pasuk in chumash is a chiddush. As the gemara itself says, “lamah li k’ra – sevara hu?” – if there is a pasuk that is needed to derive a din, it means by definition you would not know it based on sevara or common sense alone. So what do Chazal mean when they label only certain dinim as a chiddush? R' Amiel suggests that a chiddush is not measured against outside barometers of what makes sense, but is measured minei u'bei against the halacha itself. The reason eid zomein is a chiddush is because it flies in the face of the halachic rule of trei u'trei that says contradictory testimony of pairs of witnesses result in a draw. The halacha has to carve out an exception to its own *internal* set of rules and invent this new category called hazamah to avoid that conclusion. The machlokes Abayei and Rava boils down to a debate over to what degree hazamah is really just another flavor of trei u'trei (in which case it is a chiddush) and to what degree it is not. That machlokes has nothing to do with the "ta'amah d'kra" (if you will) explanation of how we can make sense of this parsha of hazamah relative to external sets of rules (our own sense of justice), which is the issue the Rishonim are addressing.
This is a thumbnail sketch of a complicated sugya and I've greatly oversimplified the machlokes Abayei and Rava (Abayei may in fact agree that eid zomeim is a chiddush, but simply disagree with the ramifications Rava draws from that in the sugya). That's a shiur for another time.