Pushing large commits or cloning large repositories over HTTPS fails with timeout
This is issue https://github.com/gitlabhq/gitlabhq/issues/3882
When committing or fetching/cloning a lot of data via HTTPS over slow networks, the process fails unless it finishes faster than the timeout from settings, even if data keeps flowing.
This never happens on GitHub, but Unicorn used by GitLab seems to cause problems and simply ignores the fact that data is still flowing and stops with a timeout. (I was able to avoid the problem to some extent by setting the timeout to 30 minutes, but this doesn't scale and no matter how high the timeout is set, it might still fail for some users.)
(See the original report for potentially more details.)