Relevant metrics have remained normal for three full days and we consider the incident resolved.
The root cause was a recent minor version library upgrade that included performance enhancements, but also an unexpected side effect on how we handled reconnections, particularly when a computer wakes up from sleep. Our team was able to adjust for the upgrade change, so reconnections would again be immediate, resulting in faster post saving like before.
