React Summit Amsterdam

Modern React applications are no longer just UI layers. They sit at the boundary of a distributed system, coordinating APIs, third-party services, feature flags, client-side caches, async rendering, and user-critical flows that can fail in several ways, some loud and others not so much. This talk looks at resilience from the frontend outward. Not in a “self-healing systems” sense, but in the practical sense of baking better guarantees into the way we design React applications, so they fail more predictably, recover more safely, and surface the right signals when things go wrong. Using real production lessons from Smallpdf, including a 35-day region-specific payment failure that went undetected while most metrics looked healthy, we’ll examine why frontend resilience is often less about preventing total outages and more about containing partial, conditional, and business-critical failures before they become expensive. We’ll explore how React architecture directly shapes recovery speed, debugging precision, and operational confidence, especially at scale. Along the way, we’ll cover the frontend as a distributed system boundary, why platform uptime does not equal successful user outcome, how to use component trees, Suspense boundaries, and error boundaries as failure-containment tools, and how feature flags can either improve safety or quietly multiply risk depending on how they are structured.