We have identified several hosted deployments where tasks are getting stuck in the queued state and failing. This is happening the worker autoscaler is having issues with scaling up workers to run tasks. We are investigating the cause of this.