Flaky test failure in blacklist(?) - investigation needed

[ghost]

There appears to be a flaky build failure in the blacklist test:

============================ test session starts ==============================
platform linux2 -- Python 2.7.6, pytest-4.4.0, py-1.8.0, pluggy-0.9.0
cachedir: .tox/py27/.pytest_cache
rootdir: /home/travis/build/kivy/python-for-android
collected 37 items                                                             
tests/test_androidmodule_ctypes_finder.py ...                            [  8%]
tests/test_graph.py ..........F.................                         [ 83%]
tests/test_logger.py .                                                   [ 86%]
tests/test_recipe.py ...                                                 [ 94%]
tests/recipes/test_gevent.py .                                           [ 97%]
tests/recipes/test_reportlab.py .                                        [100%]
=================================== FAILURES ===================================
________________________________ test_blacklist ________________________________
    def test_blacklist():
        # First, get order without blacklist:
        build_order, python_modules, bs = get_recipe_order_and_bootstrap(
            ctx, ["python3", "kivy"], None
        )
        # Now, obtain again with blacklist:
        build_order_2, python_modules_2, bs_2 = get_recipe_order_and_bootstrap(
            ctx, ["python3", "kivy"], None, blacklist=["libffi"]
        )
        assert "libffi" not in build_order_2
>       assert set(build_order_2).union({"libffi"}) == set(build_order)
E       AssertionError: assert set(['android...yjnius', ...]) == set(['android'...penssl', ...])
E         Extra items in the right set:
E         'genericndkbuild'
E         Use -v to get the full diff

My guess is that get_recipe_order_and_bootstrap sometimes picks another bootstrap as default even though it shouldn't. I'll have to look into it, I'm making this ticket mainly so I don't forget