Consider cloud providers like AWS / Azure.
Various hosted services are consumed by huge number of customers.

1. Design a system to monitor functioning of the services and baselining the expected behavior.
2. Detection of a bad/faulty version of a service.
3. Upgrading the bad service of the service without impacting the customer.
4. Testing the new version with least impact on the customer or minimum number of customers.
5. Upgrading the new version of the service across data-centers, if the new version is working as expected.

Also, outline how the baselining mechanism for the service would work?

Along with the suggestions on the problem, I would also like to get following feedbacks -

1. For an application/service developer, how relevant is the knowledge about management and orchastration of the services? Isn't this question more suited for experience on Ops side?
   Maybe, it's become an industry norm to be fully aware of DevOps?

2. Also, please provide your marking for the problem - easy / medium / hard.