So now that I am doing this myself rather then listening to some one else explain it to me. Here is whats happening.
We have 2 pools. When I Live migrate from pool01 (which is running an xcp 7.6) to pool02 running xcp 8.2.1. We do the migration with the Migration Wizard. Ever Live migration like this sends the vm to the pool master on pool02. And when the pool master exceeds CPU or Memory utilization the migration will fail and send it back.
Should it not Live migrate from pool to pool to other pool members not just the pool master?
Also, once migrated to the pool master I can then Live migrate again to another host in the pool just fine. So its not a resource issue.