This was in 2012. Let's just say that a router config only getting partially applied (because of a software bug) can cause a LOT of traffic to suddenly go places it's not supposed to go.
There's also the one where all the frontend servers worldwide went into a crash loop from a bad configuration push. The SRE doing the push noticed some "weirdness" and rolled back even before the full scope of the issue was known. That one's in the SRE book.
0. https://landing.google.com/sre/interview/ben-treynor.html