When running the performance subcommand of the CLI, using multiple threads, a portion of the tests fail. It appears to be directly related to the number of threads being used.

With 1,000 runs, with 2 threads there are only 2 failures. With more threads it seems that nearly half fail.