I'm looking over my algorithm and you're comment and I'm wondering if it would be better to do the following:

sub main_loop {
  while(
    schedule_count()
    and $hit_count < $hit_limit       #could be commented out
    and time() < $expiration
    and ! $QUIT_NOW
  ) {
    #yield();
    my $parser_thread1 = async { process_url( next_scheduled_url() );}
+;
  }
  return;
}
[download]

... rather than have the 20 static threads that loop forever, each thread would exit like the mysql threads which seem to be working fine