Heres the run down - I have Hyper-V 2012 Server configured on a 4 node cluster with 1GB iSCSI storage and NIC team with 4 physical ports and a vNIC created for each MGMT, LM and CSV.
The issue I had was that during a live migration, of only 1 guest, it would get to the 50% range and start dropping packets until a little after it had finished migrating, so around 10 pings. But not only was the guest that was migrating dropping packets, all the other guests on the source host would also drop packets, about 5-6 pings, during the live migration of that single guest.
I read thread after thread, rechecked all the configurations (I was positive it was working perfectly prior to migrating all my guests from 2008R2) but I couldn't find anything definitive.
Just the other day I added another host to this 4 node cluster to bring it up to 5. Everything was configured correctly and when I added the host, all was hunky dory. The Cluster Validation was fine and all looked good, until I moved a test VM onto the host and i found that the guest had zero connectivity to the LAN.
Again, went through everything, all seemed to be ok so I decided to rebuild the NIC team which all the vNIC sit on - After I did that, voila, it worked fine and the interesting part was when I moved a couple of test VM's onto the the new node, I wanted to test the live migration times, and wouldn't you know, the issue with the dropping packets during an LM didn't occur (only 2 pings dropped at the end of the LM as per normal) the other VM sitting on the host didn't drop any packets at all.
So it appears the issue was the NIC team, I didn't want to have to rebuild all the NIC Teams on all the hosts as this would require a fair bit of downtime, so instead I opted for testing the NVSPBIND tool (NVSPBIND Home) to unbind the vms_pp protocol and then re-bind it. It worked.
Effectively all I had to do was run the following command
nvspbind /u VM_Team (VM_Team is the name of the NIC team I have)
nvspbind /r VM_Team
The above commands will drop all network communications, I suggest moving all guest's of the host you are doing this to and also running it via iLO/iDRAC etc.
After this, LM was working perfectly again.
The only issue I still have is that I don't know why it happened or what caused it.
Anyway - Just thought I'd share as it was a unique scenario which I wasn't able to find anything definitive on.