[en] Software ecosystems can be viewed as socio-technical networks consisting of software packages developed and maintained by communities of contributors. Ecosystems evolve over time through changes in the code and the social structure, and some of these changes that may have an important impact on the sustainability of the ecosystem. Some social changes may lead to a technical degradation of the ecosystem, e.g., by resulting in abandoned software packages that are still being used by many other packages in the ecosystem. To avoid this, it is important to identify those factors leading to an increased probability of developer abandonment. Using the statistical technique of survival analysis, we empirically analyse such factors for the RubyGems software ecosystem. To achieve our goal, we analysed the development activity of gems in GitHub, as well as the social interactions between gem developers in both GitHub and developer mailing lists. Our findings showed that: the more intensive and frequent the communication, the higher the probability of remaining active longer; developers with a large num- ber of commits and multi-project activity stay longer in the ecosystem; and developers that frequently abandon projects in the ecosystem will abandon the ecosystem altogether sooner.