According to Landauer’s principle, any non-reversible system can be made reversible -that is, capable of undoing its actions- by keeping information about the past of the computation. In the area of concurrent and distributed systems, this often takes the form of memories. Memories are special devices that keep track of past states of a system execution. Memories can be looked up to restore past states, upon necessity. This paper investigates and lays down ideas on how to achieve reversibility in systems that are subject to events that, as a side effect, erase some memories, creating then holes in the structure of memories. The chosen application area is concurrent and distributed systems, where the events erasing memories are the failure of nodes.