Intermittent test failures

The Derby filesystem end-to-end tests sometimes randomly fail with a "Database error: No existing connection" error during a job status wait. Not sure what's happening here, but they succeed much of the time. There's not much of a hint beyond the stack trace. The message seems to be coming from Derby, and may be the result of a too-short wait time limit, a race condition in the test itself, or something else entirely.