Starting around 3:30pm PST, we had an incident where an unhealthy API server was in rotation in our load balancer. During that time requests routed to that server responded with a 502 which could have resulted in the clients dropping notices.
At around 5:30pm PST we removed the unhealthy server from rotation.
7:22pm PST edit: Reworded message to describe that notices are not queued on the client, but dropped when the server responds in an error state