I personally use Python over bash in Linux environments. I still find it simple yet more powerful because there are modules available that enhance the ETL pipeline in many ways, but most of all, rapid development.

The real power comes from distributing ETL workloads across multiple machines that are completely open source and tied together with queuing modules. Your ETL becomes that massively parallel processing (MPP) and data starts to streamline greater than what you can achieve with SSIS at the cost of more complexity and difficulty in management.