After the devastating forest fire in the Swedish region of Västmanland in 2014, numerous investigations and evaluations suggested measures to improve the Swedish crisis management and preparedness. Yet, after a new wave of severe forest fires in 2018, the Swedish Civil Contingencies Agency (MSB) concluded that the lessons from 2014 had not been sufficiently implemented, since several issues reoccurred. The research area of obstacles to crisis-induced learning among public organizations is rather young, and any widely acknowledged theories are still lacking. This case study focuses on the crisis-induced learning process within MSB after the 2014 forest fire. Three hypotheses are derived from previous literature and modified to the case, and tested through a process-tracing plausibility probe according to an abductive approach. The analyzed material consists of documentation from the learning process in combination with semi-structured informant interviews with current and previous members of staff. The analysis confirms that the crisis documentation was insufficient which in most cases affected the learning process negatively. It further identifies an aspect of accessibility to this obstacle which should be considered in future research. As expected, the crisis learning was mainly based on the single-loop approach, although a few indications of a deeper organizational adjustment occurred. No significant indications of conflicting opinions within MSB were found in the documents, although some informants described how incompatible opinions had emerged. In most cases they impeded the process, as expected. However, in one case the conflicting interests were perceived to improve the learning outcome. This finding suggests that conflicting opinions, in comparison to previous claims, do not necessarily prevent learning. The relation between conflicting opinions and crisis learning must thereby be further explored. Additional indications of possible obstacles were that the process depended on individuals, the institutional memory was insufficient, the learning process differed between departments, and that lacking resources prevented the implementation of measures. The study ends by suggesting learning improvements and discussing the new insights for the hypotheses which can be used in future research.