Under traditional methodologies, waterfall and the likes, companies have compartmentalized software development and platform operations into two dedicated organizational units, often with entirely different management lines imposed from the top.
This model can be continued by adding a third component into the mix - hiring dedicated “DevOps staff” to sit in-between between developers and operators.
Companies need to satisfy stakeholder demands for state of the art in-house knowledge on deployment software of the day, be it “configuration management” or “infrastructure as code” software.
Popular with this model is to keep actual infrastructure changes within operations. Development teams “provide the software”. DevOps teams “provide the infra glue”. Operations “provide deployments”. Neatly separated.
What obtains is that people writing the software are twice-removed from its effects, and that those who write the “infra glue” do not have their skin in the game during deployments either.
Combine that with on-call duties separated into yet another team (!) and you can virtually guarantee hilarity ensuing left right and center.
Software developers work very differently when their sleep is put at risk. DevOps staff will be much more diligent in ensuring everything is well tested and stable before it touches integration environments.
What seems to work much less badly: forming mixed teams, involving all three groups mentioned above, rotating on-call duties - and checklists.