Failure Recovery

Recovery starts by identifying which boundary failed.

Recovery Sequence

1. determine whether the failure happened during environment setup, data collection, report publishing, or docs build
2. inspect the tracked output tree touched by that step
3. rerun the narrowest command that proves the problem is fixed
4. review any rewritten tracked files before moving to broader commands

Recovery Warning

Do not jump straight to make app-state after a failure. That broad command can rewrite multiple tracked surfaces and make the original fault harder to isolate.

Purpose

This page explains the package-level recovery discipline for failed workflows.