Load balance ELAN's with Nexus 7K question

Hello,

First off, sorry for the length and how confusing this probably reads. I am linking a diagram that hopefully makes more sense. Though we use Nexus's in our DC's, some of the situations I get myself into get confuse me a bit to the paired nature of the cores.

https://imgur.com/a/nYC0kcF <--Diagram

I have run into a deign issue that I am not sure how best to rectify. In each of our datacenters we have a pair of Nexus 7706's in VPC. We have a single Point-to-Point circuit between two of our datacenters (DC1 & DC2)which plugs directly into Core-A at each site. There is an IP placed directly on the interfaces and everything works fine. We use EIGRP internally for routing.

Recently, we purchased several new point-to-point links from other sites to our primary datacenter (DC1) as well as a redundant p2p between. The way the ISP set it up, since so many new circuits were going to a single location, was to place an NNI at DC1 and all of the new circuits will traverse a single handoff separated by VLAN. That seemed simple enough to me. I just create an SVI at each site with the correlating VLAN and use the PE as a trunk. However, I am having an issue with EIGRP the site that has two circuits connecting each other. The new link plugs directly into core B.

Since the cores are VPC pairs, I create the SVI on both cores with their own IP and a shared HSRP IP. When the new link comes up, it establishes an adjacency with the SVI IP between the B cores, but it starts throwing the following errors reset errors for the core A SVI

23:43:06.976 %EIGRP-5-NBRCHANGE_DUAL: eigrp-100 [9341] (default-base) IP-EIGRP(0) 100: Neighbor 172.31.90.20 (Vlan102) is down: retry limit exceeded

23:43:07.048 %EIGRP-5-NBRCHANGE_DUAL: eigrp-100 [9341] (default-base) IP-EIGRP(0) 100: Neighbor 172.31.90.20 (Vlan102) is up: new adjacency

My assumption is that the route tries to establish through core B, goes to core A through the VPC link, and then A sends it through its directly connected P2P link and it fails.

One of my thoughts to try to fix is to change the first L3 P2P to VLAN 102 and put the two circuits into an ether-channel, but I am not sure that is the best play.