How would you design error handling for a large frontend app — custom error classes, the cause chain, global handlers for unhandled rejections, and preserving async stack traces?

Imagine you're building a big LEGO castle. Sometimes a piece falls off. Instead of just saying 'oops,' you want to know exactly which piece fell and why, so you can fix it. We create special labels for different problems, keep a trail of what caused them, and have a net to catch any problems we didn't expect.