Hi,
I'm guessing you meant to write "bgp session packets and bfd session packets are independent" ..?
So, from your answer (many thanks for which, by the way) there is a possible scenario whereby the BGP session could traverse one member-link of the LAG and the BFD session packets could traverse the other member-link (using the example of a 2*10G LAG) ... in which case if the member-link carrying the BGP session was intermittently flapping and transitioning up<->down<->up etc. (or the flapping link was an intermediate link and not directly connected to the router interface) the associated BFD session would do nothing to accelerate the tearing down of the BGP session because it is running over the other non-flapping member-link, and therefore the BGP session would have to wait for the hold-time to expire before being torn down?