Existing literature on affective polarization in the United States has struggled to reconcile two well-documented temporal patterns: on the one hand, affective polarization levels rise in the period leading up to elections but subsequently subside. At the same time, whenever polarization levels are assessed over long periods of time, they are found to be consistently rising. Given how strongly elections condition political affect, do they have a cumulative effect that is driving the latter result, or are they a net zero contributor? It is further unclear whether higher polarization persists longer under systems where the composition of the new government is immediately resolved after an election, as opposed to being negotiated at length. Exploring these questions requires repeatedly assessing changes in citizens' affective sentiments throughout multiple electoral cycles, an undertaking that under normal circumstances can take many years, likely accounting for the lack of evidence on the subject. Here, we report results that directly inform these questions, obtained using an original ten-wave panel study fielded in Israel during the two-year period of 2019-2021, in which an unprecedented four national elections took place in quick succession and two different coalition governments were formed, allowing for granular, by-party assessment of affective polarization changes. We uncover a striking divergence in party-specific affective polarization patterns that is conditioned by electoral performance and the composition of subsequent governing coalitions, alongside large country-wide fluctuations. We propose an integrative theoretical account of how electoral cycles contribute to the intensification and diffusion of affective polarization.