This paper presents the findings of a study combining sentiment analysis and topic modeling on Instagram posts published by members of the German Parliament, with a particular focus on the government crisis in the fall of 2024. The analyzed dataset – collected and curated for this study – contains more than 480,000 posts, including an annotated subset of 2,000 posts for evaluation. For sentiment classification, we fine-tuned the German language model GBERT-base, and employed BERTopic to discover thematic structures within the corpus. We conducted a detailed temporal analysis around the period of the coalition’s collapse, closely integrating sentiment trends with topical developments. This case study demonstrates the potential of combining sentiment analysis and topic modeling for distant reading and the large-scale exploration of political communication on social media platforms, showing that these two methods combined can be used to identify crucial events such as incidents regarding the government crisis, the Magdeburger Christmas market attack or news about the case of Gisèle Pelicot.