The following run of format-stress-smoke-test timed out on a PPC:
The test runs batches of 8 format jobs. It appears the number and duration of runs doesn't match the timeout for the task. The tests were aborted and the hang-analyzer ran on 8 different format jobs. A brief survey of the stacks looks like they were all making progress when the test timed out. We should either change the timeout or the number of runs so that this doesn't happen.