When we first started to run the tests in parallel we had less tests and tests did no required to start a full stack with distributed Tuleap instances.
Nowadays our CI pipeline running tests still run everything in parallel which means we start, at the same time and among other CPU and disk intensive tasks, 9 instances of Chrome, 7 instances of MySQL, 5 instances of nginx and PHP-FPM.
Running all the tests suites in parallel is a bit inefficient, some test suites are quick enough to be played sequentially while a longer one runs so there is no need to try to do everything in parallel. Also since it generates an high sudden load, it slows tests that should be quick so they sometimes timeout for no particular reason. Reorganizing the tests to run some of them sequentially is probably going to help to get a more stable CI test pipeline.