Testing the Untestable: A Realistic Vision of Fearlessly Testing (Almost) Every Single Accelerator Component Without Beam and Continuous Deployment Thereof

Whenever a bug in some piece of software or hardware stops beam operation, loss of time is rarely negligible and the cost (either in lost luminosity or real financial one) might be significant. Optimization of the accelerator availability is a strong motivation to avoid such kind of issues. Still, even at large accelerator labs like CERN, release cycles of many accelerator components are managed in a "deploy and pray" manner. In this paper we will give a short general overview on testing strategies used commonly in software development projects and illustrate their application on accelerator components, both hardware and software. Finally, several examples of CERN systems will be shown on which these techniques were or will be applied (LHC Beam-Based Feedbacks and LHC Luminosity Server) and describe why it is worth doing so.