Seems like quite on oversight. For most of the better pthread implementations, pthread_mutex_init() on an already 0-initialized memory block is probably a no-op, but of course we should do things correctly. Also could setup analysis tools.