I'd propose a somewhat automatic congestion toll: the toll for a segment of road adapts to keep (assuming only car traffic; extension makes it less trivial but

I'd propose a somewhat automatic congestion toll: the toll for a segment of road adapts to keep (assuming only car traffic; extension makes it less trivial but by no means impossible) the road just about empty enough to let everyone drive at the design speed (AFAIK 90 mph for most of the US interstate network). The effect is that the mechanism prevents significant latency degradation from demand. While it's not necessarily possible to go that far over night, looking at how much of the traffic/congestion is from cars, and how many occupants they have, should allow to calculate what reduction perfect car pooling (with, say, 4 people per car) would allow, and at what speed that reduced number of cars could drive (without tailgating). Then, add some safety margin and set that reduced speed as the target for the auto-toll-pricing. If car pooling physically can't fix the congestion, it'd take a bit for the problematic roads to get enough (mini)bus service to enable the full design speed. WFH isn't the solution, and proper mass transit takes a while (5-10 years) to build even if you're in a hurry.