> My un-educated answer: it depends.  If the BGP route is the next hop
> from the machine sending the ping, then yes, you'll probably need to
> restart.  The ping instance will continue to want to use the broken
> path as its next hop.  If, however, you're downstream from the BGP
> route (another hop between you and the BGP router) it shouldn't matter
> because your machine's next hop wouldn't change.
>
> Does that describe the behavior you're seeing?


Specifically, this is what we have going: A couple school districts use 
Sonicwall firewalls. The fail-over system does a sustained ping to 
responder.sonicwall.com. The path it takes 100% of the time always hits 
northernlights.gigapop.net (146.57.252.194) and another dozen hops after 
that. The state has been having a lot of difficulty with this hop, 
because of some major routing changes made by the U of M for traffic 
destined to Chicago. Almost all of our districts never know there is a 
problem, because BGP on the state routers finds a better path. However, 
the two districts  (both use sonic) sustained pings insists on using the 
northernlights hop, thus triggering their fail-over circuit, even though 
there is not an actual Internet outage. Just trying to come to a 
definitive reason why this occurs so I can possibly find a fix .