"0 successors, FD is Inaccessible" yet still in the routing table?

I ran into an issue today that I've never seen before. We lost connectivity with a remote site today. This remote site has an active/passive setup with a primary circuit and router that connects back to HQ over an L3VPN, and a backup P2P GREoIPSec tunnel. Anyway, the primary circuit goes down, and then the fun happens; because failover never occurs. Our setup looks something like this:

Remote Site Primary Router -> (L3VPN Cloud) -> BGP-peering edge router [redistribute into EIGRP] -> EIGRP Distribution Router

Throughout this whole process the Distribution router keeps routing the prefix to the edge router, even though the edge doesn't have an entry in its BGP table for this prefix anymore. I checked the EIGRP table on the distribution router and saw this:

P <prefix> 0 successors, FD is Inaccessible

I ended up clearing the neighborship; then, voila, the distribution router failed over to using the prefix actually being advertised by the backup router. Has anyone ever seen this before?

EDIT: My research suggests that when "0 successors, FD is Inaccessible seen" the routing table would be using the route from another protocol, but in this case, the distribution router in question only ran EIGRP, and the routing table was still pointing to the BGP edge router as if it were the EIGRP successor.