1) THE NATURE OF THE HUSBAND'S "HAFARAH"
QUESTION: The Gemara quotes the Mishnah in Nazir (23a) which states that a woman who accepted upon herself Nezirus and then drank wine and defiled herself with the Tum'ah of a corpse receives Malkus. The Gemara in Nazir (21b) asks a question about the nature of the husband's Hafarah and attempts to answer its question from that Mishnah. Does the Hafarah of the husband annul the Neder retroactively ("Meikar Akar") or does it merely repeal it for the future, from now on ("Meigiz Gayiz")? The Mishnah there cannot be discussing a case in which the husband did not annul the woman's Neder of Nezirus before she transgressed, because in such a case it would be obvious that she receives Malkus and the Mishnah would be unnecessary. It must be that the Mishnah is discussing a case in which the husband indeed annulled her Neder of Nezirus, but his Hafarah is effective only from now on ("Meigiz Gayiz") and not retroactively. Hence, the woman receives Malkus even though the husband later annuls the Nezirus.
The Gemara there dismisses this proof. It explains that the Mishnah actually discusses a case in which the husband did not annul the Nezirus, and indeed the Mishnah is not teaching anything new. It mentions that case only as a preamble to the case in the Seifa ("Agav").
In his explanation of the Reisha of the Mishnah in Nazir, the RAN (beginning of 83a) cites the Gemara's explanation that the Reisha is merely "Agav" the Seifa. The Ran apparently understands that the Gemara here follows the opinion that the husband's Hafarah uproots the Neder retroactively, "Meikar Akar," because that is the conclusion of the Gemara in Nazir when it says that the Reisha is "Agav" the Seifa.
Furthermore, when the Gemara here attempts to prove that one does not bring a Korban for a half-Nezirus, it cites the case of a woman who accepted Nezirus upon herself after she had already designated an animal as her Korban. Subsequently, her husband annulled her Nezirus. The woman is no longer obligated to bring the Korbanos for her Nezirus, except for the Chatas ha'Of. The RAN and ROSH explain that in that case, the woman had already become Tamei before the husband's Hafarah (because only a Nazir Tamei brings a Chatas ha'Of). This is additional proof that the Gemara here follows the opinion of that the husband's Hafarah works retroactively ("Meikar Akar") because, as the Gemara in Nazir clearly states, according to the opinion of "Meigiz Gayiz" the woman must bring all of the Korbanos of a Nazir Tum'ah if she became Tamei before the Hafarah, since the Hafarah only annuls the Neder from the time of the Hafarah and onward.
However, a closer examination reveals that the Gemara here does not follow the view that the husband's Hafarah is "Meikar Akar." The Gemara compares the case in which the husband annulled the Nezirus after the woman became Tamei to a case in which the husband annulled the Nezirus only with regard to the wine, and it calls both cases "Chatzi-Nezirus" (half-Nezirus). The Gemara's statement is consistent only with the opinion that the husband's Hafarah is "Meigiz Gayiz." According to the opinion that the husband's Hafarah is "Meikar Akar," the Nezirus is uprooted retroactively and there is no Nezirus at all, not even a "half-Nezirus." (See Ran, DH Amar Lei Abaye.). (See KEREN ORAH here, and BIRKAS ROSH to Nazir 22a.)
ANSWER: It must be that the Sugya here disagrees with the Sugya in Nazir and is of the opinion that even if the husband's Hafarah is only "Meigiz Gayiz," the woman does not bring Korbanos of a Nazir Tum'ah if she became Tamei before the Hafarah, because there is no Korban for half a Nezirus, as the Gemara here teaches. (D. Zupnik; see following Insight.)
2) WHY DOES THE PRINCIPLE OF "HUTAR MIKTZASO HUTAR KULO" NOT APPLY TO THE HUSBAND'S "HAFARAH"
QUESTION: The Gemara discusses several cases of a Neder which the husband may annul in part, due to the fact that one part involves Inuy Nefesh and the other does not. Why, though, is the entire Neder not annulled because of the principle of "Neder she'Hutar Miktzaso Hutar Kulo" -- when a Chacham annuls part of a Neder, the entire Neder becomes annulled (66a)?
(a) The ROSH (82b) explains that there is a fundamental difference between the Hatarah of a Chacham and the Hafarah of a husband. The Hatarah of a Chacham is empowered to render the Neder null and void retroactively. The entire Neder is completely uprooted from its beginning, and thus it is not possible for the Chacham to annul only part of the Neder. In contrast, the Hafarah of the husband is empowered to repeal an existing Neder from now on; it does not work retroactively. Hence, the husband is able to annul part of the Neder from now on without affecting the rest of the Neder.
This answer seems consistent only with the opinion that the husband's Hafarah is "Meigiz Gayiz" -- the husband's Hafarah works from now on and not retroactively.
(b) The RAMBAM (Hilchos Nedarim 13:10) rules that the principle of "Neder she'Hutar Miktzaso Hutar Kulo" does not apply to the Hafarah of the husband and thus the husband can annul part of a Neder. However, the Rambam also writes (13:2) that the husband "uproots the Neder from its source." This implies that the Hafarah of the husband is retroactive. If it is retroactive and uproots the Neder from its inception, why can the husband annul part of the Neder? The husband should have the same restriction as the Chacham!
The KESEF MISHNEH points out another problem with the Rambam's ruling. The Rambam seems to rule that the husband's Hafarah is "Meikar Akar," it uproots the Neder retroactively. The conclusion of the Gemara in Nazir, however, is that Hafarah is "Meigiz Gayiz"!
The Kesef Mishneh answers (as explained by the LECHEM MISHNEH) that although the Rambam writes that the husband "uproots the Neder from its source," he distinguishes between Hatarah and Hafarah. The mechanism of Hatarah renders the Neder as though it was made in error. Hafarah, in contrast, accepts the validity of the Neder but then uproots it. The Lechem Mishneh explains that this is indeed the Rambam's definition of "Meigiz Gayiz": although the Hafarah works retroactively, it nullifies an existing Neder (such that after the Neder has been nullified, the Neder is still considered as having existed), in contrast to "Meikar Akar," which -- like the Hatarah of a Chacham -- renders the Neder null as if it had never been made.
The Kesef Mishneh's explanation of the Rambam provides another way of understanding why the principle of "Neder she'Hutar Miktzaso Hutar Kulo" does not apply to the Hafarah of a husband. The Hatarah of a Chacham works by rendering the Neder as if it was made in error, and therefore it can work only on the Neder in its entirety. Although Hafarah also works retroactively, it is based on the right of the husband to nullify the Neder without uprooting it as an error, and therefore it can work even on a part of a Neder.