I can't access the article from here, but back in 1960 or so (from an article in New Scientist back then, there was a study of optimal traffic flow in a new, long, toll tunnel.
The research then indicated that there is a natural 'modal flow rate' for any lane (dependent on a variety of things -such as lane width, obstacles, etc.). This could best be expressed as a number of vehicles per hour.
When they adjusted the controls on the toll booths to allow no more than the correct number of vehicles per hour to enter, traffic flowed perfectly, except for accidents, of course. If more than the modal rate was allowed to enter, sooner or later someone would have to 'brake check', then the car behind them would 'brake check', etc. The first car would resume speed, but a little knot of between 3 and eight vehicles (a 'modal node') would move back down he lane at about 2 to 3 miles an hour. Only a gap in traffic in the lane sufficent that one vehicle did not have to 'break check' could eliminate the 'modal node', otherwise it would propagate all the way back to the beginning of the lane. If 'modal nodes' occurred, the throughput was always less than the 'modal flow rate'.
The particular solution for that tunnel was, as I recall, to allow six vehicles (one from each toll booth) in, then allow a gap of at least eight 'vehicle units' before allowing the next group of six vehicles in.
