We've reacted to an alert about queue threshold levels on the outbound network inside of our redundant outbound nodes. Upon investigation, it appears the server stopped processing new items and kept them in the "pending queue" until manually rectified. We're making adjustments to our monitoring solution to better track the age of messages that pass through the network so we're alerted not only by queue and throttling events but also this rare case of messages remaining in the pending queue.
The issue appears to have held on to random messages that arrived to outbound3 between 3am and 9am and was fully resolved by 9:45am. We've observed all pending mail was delivered.