Heads up: drafts only, for now. The titles below are real lessons from real production incidents — fleshing them out one at a time.
A post-fill margin check flipped a session-wide kill switch — and never unflipped it. Every scanner went silent while the kicker kept printing "Scanning…". The fix is twelve lines; the bug took an afternoon to find.
A one-way kill switch, a phantom 2/3 slot, and the post-incident fix that auto-unfreezes when margin recovers.
Verify-before-commit, per-host rollback, and writing the new private key back to the AWX credential store without a manual click.
A walkthrough of a self-hosted observability stack that costs nothing to leave overnight, because it doesn't get left overnight.
Putting Claude in front of every entry decision, the latency budget that survived, and the kinds of skips that actually save money.
A short note on starting this notebook, and the kinds of writing I want to put here.