Running an incident from detection to resolution, and writing a blameless postmortem that actually prevents a repeat.
Reference Q&A for this concept is coming soon.
