A data center VLAN flap resets EIGRP neighbors at dozens of branches

The company has 60 branches on DMVPN. Each branch router has two tunnels: Tunnel0 to HUB1 (172.16.0.0/24) and Tunnel1 to HUB2 (172.17.0.0/24). Everything runs classic EIGRP AS 100. The hubs send the full routing table to every branch, and no branch is configured as a stub.

At 09:12 the server VLAN 10.50.8.0/24 on DSW1 in the data center flapped for a few seconds. For the next two minutes, users at more than 20 branches lost every application, not just the servers in that VLAN. HUB1's log is full of lines like this:

%DUAL-5-NBRCHANGE: EIGRP-IPv4 100: Neighbor 172.16.0.27 (Tunnel0) is down: stuck in active

You are on the incident bridge. The same thing happened last month.