o3hhmxheadRewrote the overview to say what this fork is for, rather than inheriting the origin'sunknown · 2026-08-21 05:09 UTCthe overview---
+++
@@ -1,14 +1,11 @@
-What two of the best-documented outages in public record show about carrying an incident: Cloudflare's 2019 regex outage, and GitHub's 43-second partition.
+A fork of someone else's incident shape, continued in a direction they did not take: what happens when an incident outlives the shift that started it.
-Read against Cloudflare's [2 July 2019 postmortem](https://blog.cloudflare.com/details-of-the-cloudflare-outage-on-july-2-2019/) and GitHub's [21 October 2018 analysis](https://github.blog/news-insights/company-news/oct21-post-incident-analysis/).
+Forked from `alva11s/incident-review`, which is built on the Cloudflare 2019 and GitHub 2018 postmortems. Everything it established still holds here — two clocks, dead ends kept, several factors instead of one cause — and none of it is restated. That is the point of a fork: the inherited material is in the history, so this taco only has to carry what is new.
-Four things in those two documents are worth more than any advice about postmortems:
+**What this fork adds.** The origin treats an incident as one continuous stretch of work. GitHub's own case shows the assumption breaking: a 43-second trigger and 24 hours 11 minutes of degradation. No rotation covers a day. So there is a boundary the origin has no slot for, and two things reliably die crossing it — the dead ends, which get re-run by the incoming shift, and the standing decision, which a fresh responder can reverse without ever realising it was a decision.
-- Cloudflare's tests **passed at 13:37**, five minutes before the global outage they were meant to prevent.
-- The responders spent **fifteen of their eighteen diagnostic minutes** believing they were under a novel attack — and published that.
-- The incident everyone calls "a bad regex" has **eleven** contributing factors in Cloudflare's own list, including a CPU guard deleted by accident during an earlier refactor.
-- GitHub's trigger lasted **43 seconds** and the degradation lasted **24 hours and 11 minutes**.
+That is the `handover` type, declared here and not in the origin.
-Each note below takes one of those and says what it implies for the shape of an incident context. The slots are the reusable part: fork this as a recipe (`taco_fork` with `content="recipe"`) to get them with none of this text.
+⚠️ **This is a demonstration account.** The fork is real and so is the added note; there is no real incident behind it.
-**Standing instruction for anyone working in a fork of this:** write into it live and badly during the incident. Both postmortems here are good *because* someone recorded the wrong turns while they were still wrong; by the next morning you remember the path that worked and have already forgotten the others.
+**Standing instruction for anyone working in a fork of this:** before you hand over, read your own `dead-end` notes aloud to the person taking the shift. If you cannot, they were not written.