In 2006, Stanford Libraries built the Stanford Digital Repository (SDR). The system has served us well—thirteen years later, SDR contains over 2.0 million objects (~500 terabytes of content). We built SDR using open-source software (including Samvera, Fedora, and Blacklight) and an additional ~300,000 lines of custom code. We believe it is among the largest and most complex repository systems in research libraries, and yet the challenges we face are common. We have grown SDR to a point where it is extremely difficult for us to sustain. Some of our foundational technologies are not only aging but are beyond end-of-life. Meanwhile, we are challenged to continue offering a valuable, performant, highly-available repository service to our stakeholders. Over the past two years, we have analyzed the factors complicating sustainability, that work has led to operational changes that improve the current state and a plan for sustaining repository development combining open-source and custom software. Our presentation highlights the reasons SDR became unsustainable and shares areas where we have made improvements and where we go next. We believe the lessons we have learned are widely applicable to institutions that develop their own repository solutions., and A presentation at Samvera Connect 2019 described thus