Musk's 'significant backend' rewrite took Twitter down for five hours

On 28 December 2022, Twitter suffered a widespread outage that began around 4pm Pacific time and lasted more than five hours. Tens of thousands of users reported strange error messages, blank pages and unexplained sign-outs; many could not see replies, respond to tweets or follow trending topics, and some were shown 'rate-exceeding limit' errors suggesting servers could not cope. DownDetector and NetBlocks confirmed reports from the UK, Canada, Germany, Italy and India, with TweetDeck also down; NetBlocks said the incident was unrelated to country-level internet disruptions.

The proximate cause was self-inflicted. Musk tweeted that the firm had rolled out 'significant backend server architecture changes' that should make Twitter 'faster'. It came ten days after he tweeted that the service was operational 'even after I disconnected one of the more sensitive server racks', and weeks after Twitter briefly blocked traffic from about 30 mobile carriers, mainly in the Asia-Pacific region, in an attempt to rid the service of spam.

The context made the failure legible: Musk had acquired Twitter for $44 billion in late October 2022 and sought to cut expenses by eliminating thousands of employees, many of whom maintained the service's infrastructure, while personally pushing to remove 'bloat code' for speed. Engineers mocked the framing — software architect Grady Booch quote-tweeted: 'Architecture. You keep using that word. I do not think it means what you think it means.'

Five-hour outage hitting tens of thousands of users across the UK, Canada, Germany, Italy and India; #TwitterDown trending

The outage followed the elimination of thousands of infrastructure jobs — a rewrite landing on a network with fewer people who understood it.

It was the second self-inflicted incident in a month, after the 30-carrier blocking episode and the 'sensitive server rack' boast.

The failure mode was visible to users — rate-limit errors, blank pages, forced sign-outs — turning a performance project into a trending hashtag.

The infrastructure you don't understand is load-bearing. Performance rewrites and cost cuts meet reality the same way — especially when the people who knew the system are gone.

Twitter resolved the issue after more than five hours; the company disclosed no root-cause detail beyond the architecture-change tweet, and no user-impact figures. The episode became part of the public record of Twitter's post-acquisition instability, cited alongside subsequent service wobbles as evidence that infrastructure maintenance capacity had been cut too deep.