The backlog was cleared and performance has returned to normal.

An application has interacted with the API in such a way that has flooded the background queues.

Their rate limits have been cut to prevent the problem growing, we are taking remedial action to clear the queue backlog as soon as we can.