I thing you are doing a very subtle mistake which will become fatal in long-term. Your strategy to take small steps that cover as much functionality as possible is reasonable, but it is necessary to be careful, as it leads to a critical state when there is too much little stuff built up without proper structure to support it.

When the relations are implemented in the right way, they will simplify Gitlab, not make it more complex.

That helps us keep GitLab simple as long as we can.

One way is to have a very simple model of the same thing like Github and let users decide on the hierarchy of things. If Gitlab really want to go one step further and give users more then its really important simplicity is not killed in the process.

If you want a queueing system that uses X, use a queuing system that uses X! Sidekiq's mantra is simple and efficient. Redis is both. Abstracting data storage is neither.

A key aspect of this library is its simplicity

Keep it simple.

Lets not extend the framework with yet another syntax

"AOO is not, and isn't designed to be, the 'super coolest open source office suite with all the latest bells and whistles,'" Jagielski continued. "Our research shows that a 'basic,' functional office suite, which is streamlined with a 'simple' and uncluttered, uncomplicated UI, serves an incredible under-represented community.