Steps to reproduce: * create a project and mirror http://github.com/ceph/ceph * use https://pypi.python.org/pypi/github2gitlab to import the ~4,000 pull requests as branches * wait for the google crawler to visit * watch the worker use all the available memory until it gets killed by the limit set in config.ru (tried with 2GB instead of the default 250MB: it takes about half an hour to get to to 2GB) GitLab 7.5.1