Our network administrator was able to find a point of bandwidth saturation between a pair of core switches that was causing a bottleneck, resulting in packet loss. Connections should now be stabilized.