A single application node was underperforming, resulting in timeouts. The affected node was removed and refreshed and reintroduced.
Our teams continue to work on strategies for reducing impact from poorly performing nodes, including recent changes to our authorization connection logic, failover logic and memory management adjustments.