If you have been paying attention to the ESXi host to which the VM's are migrated during your vMotion, do the problems generally occur with the same ESXi host? That could narrow things down.
I have seen this happen in the past with a 4.x cluster... the host does not necessarily detect a problem with the uplink on a particular host, though there is in fact an issue. The VM is moved to that host, and the problem vmnic is selected to be used by the vm's vswitch port. My workaround was the same - disconnect nic, then reconnect. But upon further inspection, I found there was evidence of an issue. On one of the uplinks, the CDP information was not showing, while it was showing for all the other uplinks on the same host, and for all the uplinks on other hosts connected to the same physical switch. Ended up replacing the nic, I think.