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

i've been getting some consistent feedback in mock system design interviews lately: "your design works for the happy path but doesn't handle edge cases well." this mostly comes up when discussing distributed systems. i'll propose a basic microservice setup with a database, message queue, and cache, and then they'll ask about what happens if a node fails mid-write to the database, or if there's significant clock skew affecting ordering in an event-driven flow. i'm good at the broad strokes of breaking down requirements and choosing components, but then i get stuck on the nuances of failure modes. things like handling cache-database inconsistency during updates, dealing with partial writes, or ensuring atomicity across services. does anyone have a good mental checklist or framework they use to systematically go through common edge cases for a given component or interaction? i'm trying to move beyond just the ideal scenario.