I thought that based on previous discussion, including the runner_cfg generic would be enough for VUnit to pickup a testbench to run tests on. It appears that both the entity and filename of the testbench must also meet the following rules:

def tb_filter(entity):
    """
    Filters entities with both file name and entity name matching tb_* or *_tb.
    """
    # Above docstring can show up in ui.py warnings
    file_ok = basename(entity.file_name).startswith("tb_") or splitext(basename(entity.file_name))[0].endswith("_tb")
    entity_ok = entity.name.startswith("tb_") or entity.name.endswith("_tb")

    if file_ok and entity_ok:
        return True
    return False

This was a problem for me because a testbench was named myblock_TB.vhd instead of myblock_tb.vhd.

I suggest that the scanner allow any case when filtering on tb_* or *_tb.

It might be worth considering bypassing the naming rules if the entity contains the runner_cfg generic.

Regardless, a note should also be added to the documentation that the testbench files and entities must be named in a specific way in order for the scanner to pick them up.