Think your CI/CD optimization is on point? New research suggests we might be looking in the wrong place, revealing that your pipeline likely fails far more, and much earlier, than you realize—with a staggering 5:3 pre-merge to post-merge failure rate and 15 times more pre-merge checks. This episode unpacks the concept of "good" failures (early, cheap pre-merge fixes) versus "bad" ones (late, costly post-merge disruptions), arguing that these early issues are crucial signals. We explore why the pre-merge stage, often overlooked despite its high activity, is a goldmine for low-risk, high-impact improvements to development speed, cost, and overall quality. Learn how focusing on these "good" failures can improve developer productivity and shift CI/CD strategy from merely chasing faster builds to proactively ensuring quality where fixes are cheapest and most impactful. The discussion redefines CI/CD process milestones—pre-merge, post-merge, and post-release—and highlights how the impact and accountability for failures shift across these critical phases. Ultimately, this challenges the common focus on post-merge optimization, urging a strategic shift to leverage these numerous pre-merge "good" failures as key opportunities for building robust systems.
Read the original paper: http://arxiv.org/abs/2504.11839v1
Music: 'The Insider - A Difficult Subject'