There has been considerable research on how commuting travel times respond to metropolitan area size. Alex Anas concluded that, “The data on the largest U.S. MSAs show that commute times increase only slightly with city size: the elasticity of the average commute with respect to the number of workers is about 0.1 in 1990 and 2000” (p. 146 of Brooks, Donaghy, Knaap). This means that a large city with 2-million workers and an average commute time of 25 minutes (one-way), can add 200,000 workers and the average trip time goes up by 15 seconds.
This is good evidence for the co-location of employers and employees. No "traffic doomsday". Of all the data sources on journey to work, NHTS provides disaggregate data so that metro area means as well as variances can be computed. In 2009, across the set of large U.S. MSAs, the correlation between area mean and variance was 0.82. This means that variances, likewise, are only moderately sensitive to MSA population. But variances are more sensitive to outlier values than means. This one simple finding adds to the strength of the co-location hypothesis.
People are not stupid. They (employers and employees) do what they can to accommodate to the challenges of big-city life. But traffic doomsday has constituents. You know who you are. Hint.