I'm seeing an increasing use of serverless technologies and managed services. Why nurse maid a Zookeeper and Kafka combination when you can use a managed Kafka instance?
ETL solutions such as Matillion, AWS Glue and Google DataFlow offer good and improving facilities that cater for the majority of needs.
Data warehouse in the cloud solutions that separate storage from compute have come over the horizon. Snowflake DB keynoted at both the AWS and Microsoft conferences. Microfocus Vertica EON takes a similar approach with the data stored in highly resilient S3 storage and compute nodes providing the querying facilities.
I still think being in the cloud demands a higher level of due diligence across the board. You call out security and I agree. The bills for an inefficient app or process can be scary.
I do worry about systems, allegedly for data science, that charge per query. Sooner or later the bean counters will be looking at a job role that exists to ask questions and use a financial stick to limit the questions they are allowed to ask.