What may have caused this BGP-EIGRP loop to form?

Issue: during the day, suddenly all field offices lost connection to site A. Traceroute showed the following loop:

​

1. A_RTR1 VPLS interface
2. B_RTR2 MPLS interface
3. B_RTR1
4. A_RTR1 VPLS interface

​

Diagram

​

• A_RTRs learn the 10.1.0.0/16 subnet via EIGRP AS 1 from switches
• RTRs run EIGRP AS 1 over VPLS
• RTRs run BGP over MPLS
• RTRs were mutually distributing EIGRP into BGP and vice versa.

​

What broke the loop: removed EIGRP redistribution into BGP.

​

Question: How did this just happen in the middle of the day? My thinking is that site A RTRs lost EIGRP routes to 10.1.0.0/16, then learned the route via BGP from B_RTR2. However, EIGRP neighbors between Site A RTRs and SWs was up the entire time.