Mock interview feedback: 'your design doesn't handle edge cases' — how to improve?

I've received feedback in mock interviews that 'your design doesn't handle edge cases.' My designs usually work great for the happy path, but I often overlook scenarios like a node failing mid-write, clock skew causing inconsistencies, or caching issues leading to stale data when the database updates. It's frustrating because these are the things that break production! I need a mental checklist or a systematic way to consider these failure modes during the design process. How do experienced engineers proactively think about edge cases? Is it about enumerating different types of failures (network, node, process, data) and then asking 'what if this happens?' for each component? Or are there specific architectural patterns that inherently address certain edge cases? I feel like I'm missing a structured approach to thinking about resilience and fault tolerance beyond the basic 'add a replica' or 'retry mechanism.' Any practical advice or frameworks would be incredibly helpful.