In the world of quantitative finance, software errors can trigger the loss of millions of dollars, and even brief system failures can be costly. At the same time, a firm's competitive edge is highly dependent on the speed and adaptability of its software. In this informal talk, we explore the delicate balance between reliability, flexibility, and performance in real-time trading systems. In particular, we discuss some of the key challenges of designing fault-tolerant architectures and employing error prevention and detection tactics while achieving the highest possible performance. We also cover some of the techniques used to leverage the contributions of hundreds of developers, researchers, and systems engineers while limiting the potential for problems in the production environment. The talk aims to provide some real world examples of software reliability challenges, and how they are solved in practice.