Assume every driver is equally at fault.
Each driver is causing 1000 person-seconds of slowdown.
Each driver is receiving 1000 person-seconds of slowdown.
Seems fair to me, and each driver is paying their exact cost.
It's wrong to look at the externalities drivers output and ignore the externalities drivers input.
In some situations the margin of newcomers is a fair way to look at it. But with drivers on the road, these people are all getting up in the morning and choosing to use this road at the same time. It is a continuous reaffirmation, day by day, that this is the path they want, no matter who they slow down, and they accept being slowed in return.
"Each driver is paying their exact cost." Yes, but on the margin each driver can only choose a small chunk of their cost. Seriously, have you looked at the enormous literature on this? One major annoyance on HN is the way CS people assume that they can independently grok the intricacies of any field.
Assuming drivers actually consider their effect on congestion, they are willingly accepting the congestion of those coming after just as much as they accept the congestion they cause. (Most probably don't consider it but that's beside the point) Most importantly, the drivers are choosing to use the road every day even after seeing the final congestion every day. That's not a marginal decision. They are choosing to accept the final congestion.
But even if you insist on a marginal explanation with actors incapable of anticipating future cars, the average driver is intentionally accepting and causing 500 with past cars, and unwittingly accepting and causing 500 with future cars. Versus your characterization of them only intentionally having 10 delay.
Framing it as a new car 'paying a pittance more' is nonsense. The twentieth car either doesn't travel or attempts to pay 200 to travel. Comparing 190 to 200 is meaningless, because they never had the chance to travel at 190.
There is a congestion externality on them despite not using the road.