Sanhedrin 27a - Eid Zomeim is a Chiddush

Rava holds that if one testifies in nissan and becomes an eid zomeim in tishrei, he is not invalidated as a witness retroactively from nissan and is only invalid from tishrei and on. The rationale according to the first version of the gemara is that eid zomeim is a chiddush so he would only be invalid from the time that the chiddush is introduced, namely tishrei. Rashi explains that the chiddush of eidim zomemin is that the second two who claim "you were with us elsewhere" have the ability to invalidate the first two. Tosafos struggles with this because in Baba Basra 31b, Rava himself entertains the notion that even when two groups of witnesses contradict one another, they are all invalid for other testimonies. Clearly, the sevara is that when there is a contradiction both pairs should be void. Why then is the invalidation of the first group considered a chiddush? Tosafos offers two approaches. In the second approach, Tosafos suggests that the chiddush of eidim zomemin is not that the first group becomes invalid, but rather that the second group remains kasher. The logic of Tosafos is, just as we invalidate two groups who contradict one another about an event, we should invalidate both groups when they contradict one another about the plausibility of the first group witnessing the event.

Tosafos struggles with their second approach - Since the chiddush is that the second group remains valid, but it is not at all a chiddush that the first group is passul, what is the justification of the gemara to only invalidate them from hereon in and not retroactively from the date that they testified? Tosafos asnwers that since there is a chiddush regarding the leniency of eidim zomemin that the second group remains valid, we cannot learn even from the invalidation of the first group. Tosafos draws a parallel from chulin 99 where the gemara says that we can't learn the concept of ta'am k'ikar from meat and milk because ba'sar v'chalav is a chiddush. Although the chiddush of basar b'chalav is a leniency that they don't become assur unless they are cooked together, we cannot learn from the stringency.

R. Akiva Eiger (Gilyon HaShas) asks, what is the comparison between eidim zomemin and ba'sar v'chalav? By basar v'chalav there is a chiddush in the leniency of it, so we are concerned that the stringency of ta'am k'ikar is also a chiddush, therefore we cannot learn from there to elsewhere. But by eidim zomemin, the stringency of invalidating the first group is a sevara, and not part of the chiddush of the din. Therefore, the chiddush in the leniency of the second group remaining valid shouldn't have any bearing on the first group. The invalidation of the first group is a sevara that is not sourced in eidim zomemin and applies even in cases of contradictory witnesses, so why should it be regarded as a chiddush? Perhaps Tosafos holds that since we find a chiddush regarding the leniency of eidim zomemin that the second group is valid, the din should have been that the first group is also kasher. The contrast of the first group being passul in comparison to the second one being kasher is reflective of a chiddush even in the invalidation of the first group.