The evolution of a socio-technical system is an iterative and collaborative process, in part driven by failures of the system. When a system failure is sufficiently severe, an organization may conduct an incident analysis to learn from the failure and choose post-incident action items to co-evolve the technical and human aspects of the system. Developing a better understanding of this failure-driven software evolution processes is the goal of our ongoing research project. As a first step, we have collected publicly published incident reports, extracted and analyzed 104 post-incident action items in those reports. The focus of our analysis has been on: (1) what motivates those actions, (2) what changes are being made to systems, (3) which parts of the systems are being changed, and (4) the goals of the changes. In this paper we report on preliminary findings from this analysis. Another aspect of this software evolution process that has not yet been studied (as far as we have been able to determine), is the relationship between incident analysis (along with the work that follows the analysis) and requirements engineering. So in this paper, we also discuss how aspects of requirements engineering may be helpful in addressing challenges or open questions related to incident analysis.