The main reason is to make bisect easier.
In some points, we have a lot of git dependencies. Since we don't have
the information of which commit we are referring to, bundler get the
latest commit of the master branch of the dependency. This sometimes
returns a version that is not compatible with Rails anymore, making the
tests fail and the harder to identify the commit that introduced a bug.
Also this will make sure that a contributor will always get a set of
dependencies that are passing with our tests.
In our CI server we delete the lock file to make sure we are always
testing against the newest release of our dependencies.
