Recently, communication scholars have paid attention to the growing dissonant and dissipative character of the public spheres, especially in their connection to networked discursive spaces. While substantial dissonance of the discussions is well addressed, structural discontinuity of public discussion remains under-explored. Reproducibility of the discussions on similar issues or events in time, we argue, needs to be seen as a marker of stability of public spheres. In this paper, we compare the user and influencer structure of two similar discussions on German Twitter of 2016 (the Cologne mass harassment) and 2019 (the Chemnitz killing). We show that the overall reproducibility of the discussions is extremely low, and the only structural element that reproduces are influential media, mostly of national reach. But even the stability of media presence must be questioned, as both intensity of their presence within the discussion and user engagement with their tweets varies much from one discussion to the other. Thus, one may conclude that the structural stability of public discussion of similar events on Twitter is not reached.