How does molecule combine group_vars of the same name?

[isuftin]

TLDR: Why does Molecule (or Ansible) prioritize using a playbook's group_vars contents over Molecule's when the files have the same name? How can we enforce that Molecule use the group_vars files in Molecule's configuration instead of the playbook's group_vars?

I have a playbook. The playbook has group_vars files of the same name in both the molecule's group_vars dir as well as the playbook's group_vars dir.

.
├── group_vars
│   ├── test.yml
├── molecule
│   ├── aws-marketplace-centos7-ec2
│   │   ├── molecule.yml
│   ├── aws-marketplace-rhel8-ec2
│   │   ├── molecule.yml
│   ├── common
│   │   ├── converge.yml
│   │   ├── group_vars
│   │   │   ├── test.yml
├── release.config.js
├── requirements.in
├── requirements.txt
├── roles
│   ├── requirements.yml
├── site.yml

In each scenario, the test instance is placed in the test group. The test group under the top-level group_vars uses the community.hashi_vault.hashi_vault plugin to look up secrets:

hashicorp_vault_value: '{{ lookup("community.hashi_vault.hashi_vault", "secret/data/some/location/in/vault") }}'

This works when the playbook is running on AWX/Ansible Automation Platform because AAP will feel the plugin the necessary credentials and endpoint information. When I run my playbook in Molecule, this of course is not the case, so the test group under molecule/common/group_vars/test.yml has the same variable name, but with a filler value:

hashicorp_vault_value: Not using hashicorp vault

The scenarios link to the group_vars under molecule properly:

provisioner:
  inventory:
    links:
      group_vars: ../common/group_vars

I have a task in my playbook that uses the hashicorp_vault_value to write to a Jinja2 template file.

So when I run Molecule, it fails because it tries to use the community.hashi_vault.hashi_vault plugin to do the thing. So this tells me that either Molecule or Ansible (7.1.0) is prioritizing the playbook's top level group_vars over what I'm linking in Molecule.

If I change the group_var filename under molecule's common dir to something else and change the group name in the scenario to that new value, everything works fine.

So... does anyone know how Molecule or Ansible does this under the hood and if it's possible to predictably get Molecule to use the group_vars provided FOR molecule instead of the playbook's group_vars if both have the same filename?