That is a good question, Tarun.You've hit on the core nuance: small design choices often calcify into architecture through sheer volume and gravity. To answer your question: I draw the line based on the projected blast radius at the moment the choice is made, rather than just its immediate cost. If a choice defines foundational data flow (like state management or API contracts), we must anticipate that its cost of change will compound exponentially once adopted across components. Treating it as merely 'low-cost design' early on is how accidental architecture is born. The CSS/Tailwind example works because styling layers have natural boundaries and low coupling. But for data and state, a pattern copied across 40 components isn't just design anymore—it's structural. That’s why architectural governance needs to look ahead at the compounded cost, not just day-one convenience.