A race condition exists whereby Logger::setLogFile can close Logger::fd while Logger:rintThreadMain is using it to write output to log, causing a failure. This occurs because the fd variable is not protected when used by printThreadMain.
That being said, I would consider this a very minor bug since setLogFile is only really used within OptionParser and OptionParser executes before any log messages are produced. This bug only shows up when one runs LogTests.* on a very resource-constrained VM since it rapidly calls into setLogFile between tests.
This bug is reproducible as of commit cd81885.