The main assumption in the Integration-Manager workflow is that code from repositories of other users is always pulled by the owner of the current repository as and when appropriate.

So if dev1 and dev2 are working on the same feature in 2 different forks of the main repository, dev1 has to pull commits from dev2 that are needed in his/her fork and dev2 has to do the same in his/her fork. Once the feature is complete, the merge request is created from one of the 2 forks.

Yes it is harder to mention issues, but that can be done in the message of the merge commit. Since forks are essentially equivalent to branches in this workflow, I usually don't mind referring to issues in the individual commits itself which would link to the correct issue on getting merged to the main repository.

We do this with our team's projects hosted on Bitbucket, ymmv.