Pre-built pre-assured components are the key to building most types of high reliability engineering systems at reasonable cost, but software engineering has not found a way to use this approach. Technologies exist for building software by connecting together pre-built components but this only addresses one part of the problem. The key additional requirement is to control the reliability of the resulting software programs. New statistical reliability models appear to offer the breakthrough required to achieve this. The proposal is to assess the theoretical and practical feasibility of using these models to provide a powerful new reliability assurance method for critical software.