Chris Webb shares some insight:
Recently we announced an important new optimisation for DirectQuery datasets: the ability to run (some) of the queries generated by a single DAX query in parallel. You can read the blog post here:
A few of us on the Power BI CAT team have tested this out with customers and seen some great results, so I thought I’d write a post illustrating the effect this optimisation can have and explaining when it can and can’t help.
Chris has examples of great success, as well as not-so-great success and utter failure, and explains the why behind each outcome.