Starting at 2:15pm EDT AWS began experiencing significant disruption to many of its services, including interruptions to network connectivity on EC2 instances. This caused some requests to Trello.com to take longer than usual or timeout completely. The worst of the service degradation had passed by 2:40pm EDT while some lingering isolated affects continued, which we believed were the cause of the CDN errors that some of our users were reporting when trying to load Trello.

However, on further investigation and with the help of some details from Trello users on Twitter, we discovered that there was a problem with a recent release of the web client which affected approximately 10% of users' web sessions. The issue caused affected users to load an empty version of our core.css file, which caused Trello to display without any styling and include a message indicating that the browser was "unable to load all of Trello's resources." The mobile versions of Trello were unaffected. Once the issue was identified, we immediately reverted to an older stable version.

These two issues appear to be unrelated, but identifying and fixing the second issue took longer than we would have liked due to the increased noise in our monitoring systems from the AWS service disruption, and the initial assumption that all user issues were related to the AWS problems.