Due to a bad infrastructure definition migration, our async workers that process merge jobs started failing throughout the day. By the evening, we were able to fix the configurations, re-deploy, and restore functionality.