Bartłomiej Piotrowski commented:
It turns out to be ordinary time out caused by spinning hard drives in these boxes. For the time being it would be better to change the timeout to 120s (just to be on a safe side) as the test takes between 30-80s from what I saw locally.
If gitlab-runner ever gets ability to use tmpfs for in-container /tmp, we will be also able to switch to that. The issue is tracked here: https://gitlab.com/gitlab-org/gitlab-runner/issues/1454