Follows the discussion of #5879.
As suggested by Dan Schult in pull request I should rewrite descendants_at_distance(G, source, distance) using the bfs_layers(G, sources) in a new issue. If a new issue is created I would like to discuss some design choices if possible.
My apologies of if sent this mail to soon.
Thanks,
Dhaval