Because of a teaming policy, if an uplink goes down, connections are switched to another uplink in the team. When the first one goes up again, connections are immediately restored. This can cause packet loss when a connection is not stable and goes up/down several times for example when a switch starts.
To modify the TeamPolicyUpDelay, from vSphere go to Configuration > Advanced Settings > Net > Net.teampolicyupdelay. The default is set to 100 ms.
A configuration option, Net/TeamPolicyUpDelay is added to change the threshold in milliseconds (ms). At start up, the very first up event is not delayed. This setting delays the up notification. When an up event is received, a timer is triggered to notify this. If a down event occurs when the trigger is pending, then it is cancelled.