Some customers reported symptoms of severely delayed message deliveries and slow to start meet merge operations. These issues were traced back to much slower than usual processing of meet-related notifications to mobile app subscribers. Those slower jobs were delaying other background jobs in the same queue. Automated scaling increased capacity as expected, but took some time to catch up with the backlog.
The root cause for this incident was traced back to a change deployed the previous day to expand the automatic favoriting of athletes to include all members of the same account on any team. That change, while well tested, proved to be non-performant in a production environment. That change has been rolled back, and work on a performant version to add the same functionality is in-progress using a different approach.