Yes, it’s one test run, the 21 tests gets split into 3 devices (7 per device).
I also had the same timeout issue in one of my projects, we decided to split the large set of tests into test suites. But yeah you can also split the tests using Flank, give it a try, it’s simple to integrate :)
Ping me if you have any questions, and also if you have issues with Flank you can open a git issue on their repo.
Thanks for reading,