If workers are down, received requests are waiting in a queue to be proceeded.
Recently, based on average data, it looks like this system has gone down for about 689 minutes at a time.