Possible flakey test test_init_swarm_with_ca_config

https://github.com/docker/docker-py/pull/2288 initially failed with:

=================================== FAILURES ===================================
___________________ SwarmTest.test_init_swarm_with_ca_config ___________________
tests/integration/api_swarm_test.py:60: in test_init_swarm_with_ca_config
    assert self.init_swarm(swarm_spec=spec)
tests/integration/base.py:128: in init_swarm
    return self._init_swarm(self.client, **kwargs)
tests/integration/base.py:95: in _init_swarm
    '127.0.0.1', listen_addr=helpers.swarm_listen_addr(), **kwargs
docker/utils/decorators.py:34: in wrapper
    return f(self, *args, **kwargs)
docker/api/swarm.py:129: in init_swarm
    self._raise_for_status(response)
docker/api/client.py:263: in _raise_for_status
    raise create_api_error_from_http_exception(e)
docker/errors.py:31: in create_api_error_from_http_exception
    raise cls(e, response=response, explanation=explanation)
E   APIError: 503 Server Error: Service Unavailable ("Swarm is encrypted and needs to be unlocked before it can be used. Please use "docker swarm unlock" to unlock it.")

It passed on repeat.

I would hypothesise that there is perhaps a race in the test between the "swarm init" and "swarm inspect" where the init is not fully complete.

Start with tests/integration/api_swarm_test.py, especially SwarmTest.test_init_swarm_with_ca_config, then inspect tests/integration/base.py around init_swarm and _init_swarm. Run the integration test repeatedly and investigate the suspected timing between swarm init and swarm inspect. Done means the test no longer fails intermittently when the swarm uses CA configuration.