Retries, Caching & Logging

Day 63 of 100 Days of MLOps. Yesterday your pipeline became a Prefect flow — but a transient blip still killed a step. Today you add the reliability machinery that makes orchestration worth it: retries=3 so a transient failure is automatically retried instead of fatal, caching so an expensive step doesn't re-run on the same input, and proper Prefect logging so every run tells you what happened. Three tiny changes that turn a tracked flow into a resilient one. Runs 100% locally on any OS.