Some customers have reported that they have tests that run successfully when run sequentially, but they fail when they run them in parallel.
We've investigated a number of cases like this in depth. Historically, there are three things that have caused this:
(1) The tests are not truly independent and running them in parallel causes them to step on each other
(2) The staged app is on an underpowered machine and is getting loaded down
(3) The test runner was not thread safe and the tests were stepping on each other.
While we can never say with certainty that there is not a problem on the Sauce side, our experience has shown that it's unlikely. Our wiki includes a topic on Troubleshooting Parallel Tests that should help you get to the bottom of the problem.