We catch exceptions that throw here, which means that we don't have a backtrace for these exceptions, which leads to hard-to-diagnose issues like SERVER-43021. We should stop catching these exceptions and let the process crash on its own due to the noexcept nature of lambdas passed to std::thread