All systems are likely to be affected by obsolescence. A component is considered obsolete when it is no longer manufactured or requested. Obsolescence must be taken into account, analysed and resolved effectively as it can lead to temporary or permanent stoppage of a system or make it impossible to repair it. The consequences of an obsolete entity must be minimised as they may have impacts, among others, on system characteristics such as availability or maintainability. These consequences can cascade through the system architecture. Understanding the propagation mechanisms can greatly contribute to obsolescence resilient design, but also to obsolescence management enabling a more effective determination of the scope of impact of actual or predicted obsolescence. The objective of this research work is to propose models to describe such propagation and to understand its principles. In fact, even if the obsolescence or rarefaction propagation is real and commonly accepted, there are no models describing its mechanisms. This paper focuses on two main points. First, it contributes to the concepts of obsolescence and rarefaction. Second, we propose descriptive and mathematical models of obsolescence and rarefaction propagation which allow to describe the likelihood of this propagation through the architecture of the system via two types of identified links, namely dependency and jump-up/jump-downs. These proposals are then illustrated through an example of a monitoring unit for a perfume bottle packaging line. After a discussion of the results obtained, the paper ends with the definition of future research pathways to ease the use of probabilistic inference tools for the exploitation of the proposed models.