Following this question postgresql thread safety for temporary tables my case is slightly different, what I experience is that creating uniquely named temp tables seem to be unsafe, I simply have a test that runs few times, and fails when I witness that creation of the temp table of the second thread is probably interfered by the first thread.