First, I'm going to address mis-use of intersections. Allowing cars to pull forward into an intersection that they can't clear is an example of the tragedy of the commons (http://en.wikipedia.org/wiki/Tragedy_o
Next, a more fanciful and less penal idea. To illustrate this, I need to digress for a moment and describe something analogous.
There is a computer program called Folding@Home (http://folding.stanford.edu/English/Ho
I want to apply this approach to intersection management and traffic signals. Virtualize a wide variety of traffic scenarios and let real people on the internet try to manage them as efficiently as possible. Produce ratings and scores, set goals and achievements, inspire competition and exceptional performance. Let them trigger signal changes, handle arriving vehicles and pedestrians, view statistics on the intersection, the locale, other players' performance in that scenario, etc. In other words, make it a game, with the same sort of competitive online community that other such games have. And then, perhaps without telling people when it's actually happening, occasionally give the best players control of real world intersections. The ones with the worst traffic, or the most complicated traffic patterns, or even ones that are simply up for re-programming soon for which additional handling data would be useful.
How often have you arrived at an intersection just before it turned red, annoyed at the timing all the more because there is no one yet waiting to use the newly green light in the other direction? While waiting at a red light, have you had the realization that a gap in cross traffic is more than large enough for an entire cycle of the light to let you go through without delaying the other people? Have you ever cursed when you are stuck with a red light while oncoming traffic has a green left turn arrow that they don't need? Can you imagine being part of a pack of cars released by one intersection just in time to get caught at the next? These are all situations that it is difficult to handle with relatively simple computers programmed with less than perfect traffic detection systems and heavy pre-determined assumptions. They are all situations that an average high schooler with their hands on the switchbox could solve immediately. Such a straightforward approach would certainly be ripe with potential and actual abuse, but if you implement the qualification and monitoring layer of a game community as a gatekeeper then I expect you would see exceptionally positive results with very few negative outcomes. Constraints on actions of the "players", such as hard maximum signal cycle times and wait times, would eliminate most of the feasible remaining avenues for problems.
I have little hope of seeing the second idea implemented any time soon. The first, however, is something I do look forward to. I expect that eventually some municipalities will use red light cameras to enforce existing laws against blocking intersections, but even if they don't then someone else will have to wise up to NYC's approach with letting "meter maids" write citations for these other cars stopped where they shouldn't be.