Agents flows and error handling

Following my earlier post about creating an agent flow. I'm now writing about error handling in Agent Flows. How do we handle basic failures within agent flows?