I think the lesson to learn is that you should hedge your bets. Instead of using one service, perhaps you can protect yourself against a catastrophic failure. Chaos engineering relies on these kind of ideas. If you inflict chaos on yourself, you will deal with it better when it eventually catches up with you.
When it comes to Gitlab, this would mean managing your source and related metadata so that it is possible to access without the service. I would love to see mirroring between services as that would avoid the overall risk. This would be possible especially with Git. It's issue tracking and the rest that's harder to manage at the moment.