I wonder if Boost-timed locks have a better precision. Because then you could just lock on a mutex that nobody ever releases, and when the timeout is reached, continue on...Timeouts are set with boost::system_time + boost::milliseconds & cie (xtime is deprecated).