I remember seeing this in earlier studies of California toll lanes. The tolls are supposed to adjust to keep congestion on the toll lane down, but the problem is that the toll price is used by commuters (correctly or not) as a signal for how
congested the non-toll lanes are.
