Why does `bender` look upward for a `Bender.local` even if it is run in a Bender repo?

[FrancescoConti]

In the following situation, bender works in a way I fail to rationalize. Assume you have a nest of IPs: pulp depends on pulp_cluster (and contains it), which depends on cv32e40p (and contains it). Initally, you are working on pulp_cluster so you bender clone it and its dependency cv32e40p. This results in two working dirs pulp/working_dir/pulp_cluster and pulp/working_dir/cv32e40p. The Bender.local in pulp gets updated. So far so good.

When you finish working on pulp_cluster and cv32e40p, you push everything and update pulp_cluster's Bender.yml - but you may want to push also pulp_cluster's Bender.lock, e.g., for CI. That's when (in my view irrationally) a bender update will fail to work: it will actually point to the Path defined in the pulp Bender.local, even if the command is run inside pulp_cluster.

I understand why bender looks upwards in the file system hierarchy -- if you run it inside a nested folder in a repo, it makes sense to do that (like git). But it should stop at the first one it finds, not look further upwards -- what is the use case for that?