The postmortem I sat on
We had an incident in March. Nineteen minutes, orders lost, one customer call. Standard process: postmortem within 48 hours, blameless, action items. I wrote it in 40 minutes. Then I sat on it for a week. The failure traced back to a config I changed myself in January. Two reviewers, nobody caught it, including me. Every word in that document was true, and I still could not press send. A senior engineer once told me that blameless postmortems are for teams that already trust each other. I thought I understood him. In March I found out I did not. The process was blameless. I was not. A teammate sent his own version first. Same facts, and somehow it left out the part where I could have caught it. I read it three times, then sent mine below his. Not sure what the lesson is. Maybe honesty also has an order of operations.