We ran and still run Eigrp. We kept a simple floating static route (AD of 120) to route IP out the Bri interface should the Primary Serial Int fail. Since the introduction of BGP (MCI Private INTERNET) our routers are essentially edge routers. When the primary interface fails on one end the Backup INt (120) routes into a "black hole of BGP" on the remote end, until of course BGP (20) can figure out the route is no longer valid. Former 2 second failover is now turning into 1:20 second failover. What direction should I be persuing before impulsively grabbing the duct tape??? Bottom line BGP is not getting the word that our Remote End has failed, while EIGRP always did this instantaneously.
If the interface to the SP goes down, bgp fast-external-fallover should take the BGP session down automatically. This feature is on by default.
If the interface doesn't go down then the best way to quickly detect that the peer is not responding is to lower the BGP timers (ie: timers bgp 3 9).
Hope this helps,