True, but not relevant. Each car, tractor-trailer or sports car can move off while maintaining a safe distance from the vehicle in front.
Add me to the list of people that doesn’t consider this to be some crazy utopian dream.
In Chinese cities most green lights have a countdown on them. Now, Chinese roads are hella dangerous, but there are various reasons for that, and rear-ending while pulling off doesnt seem to be one of them. Certainly I never saw that happen in 8 years of living there.
Probably why countries like the US and UK have begun introducing them.
It absolutely IS relevant. A Porsche with Bremo brakes is going to stop a hell of a lot faster than a fully loaded tractor-trailer; therefore, the safe distance is greater behind the Porsche than being behind the TT. Unless you want to apply the Porsche distance to all vehicles, which means wasted gaps & less vehicle throughput. Just like road conditions matter; you need more distance to stop when it’s rainy or icy than you do on a dry day.
Why should all the vehicles be equally spaced? I’m not sure where you’re getting that.
Let’s take it step by step and try to find where we disagree:
- Define “safe following distance” as a sufficient length to allow you to bring your vehicle safely to a stop if the vehicle in front stops suddenly. (The exact definition doesn’t matter much.)
- The safe distance depends on various factors including speed, vehicle, driver reaction times, weather and road surface conditions.
- Other variables (vehicle, driver, road) being equal, for any speed there is a safe following distance.
- The relationship between safe distance and speed is a continuous (and monotonically increasing) function.
- Other variables being equal, for any distance (gap between two vehicles) there is a safe speed.
- A line of vehicles (including Porsches and tractor-trailers) can move off simultaneously, each increasing speed while each maintaining at all times a safe distance from the vehicle in front.
But each must move off more slowly than the one ahead, so that the gap between then opens up as speed increases. Which means that, even if they all start instantly on the green light, all but the first few cars will accelerate so slowly that, functionally, they might as well not have started accelerating at all.
The OP doesn’t say they start out with minimal space between them. They could have enough space between them while waiting at the light.
All cars and light trucks will stop well within established safety distances. 18 wheelers fully loaded may take 4 times the distance of cars to stop.
If they all queue at the lights with a space between them equal to the space they will require when travelling at the speed limit for the road then, yeah, they could then all accelerate at the same rate as soon as the lights turned green, and continue accelerating until they reached the speed limit.
But, leaving aside the question of how you would get people to queue like this, what would be gained by doing so? All you would do is vastly increase the roadspace required to accommodate queues at traffic lights, which would tend to increase congestion and/or reduce the vehicle capacity of the road in question.
Would it though? If the street can handle the vehicle capacity when moving, why would stopping in place change that? Compressing the stream of vehicles at an intersection with a red light allows vehicles to queue closer to the intersection, but it takes just as much time for each vehicle to make it through the intersection. Or in other words, the last person to make it through the light isn’t going to change, they either sit at the back of a compressed queue and wait longer until they get their chance to move, barely making it through before the light changes, or if everyone stopped where they were when the light turned red, as soon as the light turns green they start moving from farther back and still just barely make it through. Is it a 1:1 ratio? I don’t know, but it might be.
If we have networked self driving cars, do intersections even make sense? It seems like you could use roundabouts (the heavier the traffic the wider the circle can be), with cars subtly adjusting their speeds both on the roundabout and as they approach to seamlessly merge with gaps much smaller than humans could manage and without slowing down very much.
Of course even perfectly networked AI cars aren’t going to be able to drive around with their bumpers lightly toucging at highway speeds, because of imperfections in the real world - everything from road conditions to errant gusts of wind. But a safe gap would be a lot smaller than what it is for us humans.
In a “spaced” queue, everybody in the queue can start to move at the same time, but all except the first car have to travel further before they get through the intersection. If we have a road on which the speed limit is 60 kmh, the recommended car gap would be 45m. Allowing 3m for the car itself, a queue of twenty cars would require 900m of roadspace (which in many suburban contexts would exceed the distance between two sets of traffic lights, but leave that aside. Plus, unless the road is straight, people at the back of the queue may be unable to see the lights, and so won’t know when they change, which is another problem we’ll leave to one side.) From a standing start, it will take nearly a minute for the twentieth car to reach the lights. People whose priority is to get through the lights before they change are not going to stop and wait nearly kilometre away.
Traffic would flow much better, but the downside is that the line of cars moving together would only be about 1/2 car length apart so if ANYTHING caused one of the cars to have to brake, then a chain accident would result.
Agree on 1-5, but you’re not exactly right on #6. The Porsche can accel, decel, & handle curves faster than the TT can. By your having all vehicles start simultaneously, everyone is forced to the least common denominator - the speed of the TT. I can depress the accelerator of my Porsche at the same time you do in your TT or I can do that 5 seconds later & still be at the minimum safe distance by the time you get up to speed
A roundabout takes up more real estate than an intersection. Intersections would have to remain in cities & even most suburbs due to existing buildings. However, traffic lights could go away in FutureWorld® where all vehicles are Self Driven Vehicles. If vehicles could broadcast speed, speed trend (accel, steady, decel), direction, position (including lane), & length accurately (GPS has issues in urban canyons) than vehicles could seamlessly flow thru an intersection w/o stopping as demonstrated in this older religious ‘documentary’ which, obviously only the lead vehicle has devine intervention to avoid any of the other vehicles. 
Farther yes, but does it necessarily take longer? The farther back in a normal queue you go, the longer one has to wait for the vehicle in front to start moving, and then they have to span the distance from where they were waiting to the intersection itself. In a spaced queue they all start moving at once, so nobody is waiting at a green light. The converse of course is that nobody keeps moving at a red light either. I’m positing that taking out the slack and adding back the slack at every stoplight is at best neutral and may be somewhat detrimental to overall throughput.
We do have to ignore the practical questions to evaluate the theoretical one (spherical cars traveling in a straight line through a frictionless vacuum 
).
That’s fairly generous. Two seconds is the oft-quoted minimum safe following distance, which is 33m at 60kph. Regardless, it’s still a long queue, but the point is it’s no different than when all the vehicles are moving.
And if the spaced queue is a turning lane, all of the spaces will immediately filled by line cutters.
What happens if everyone goes when the light changes to green?
Lots of extra business for detailing shops.
Strange thread.
I can honestly say I have seen cars pulling away at a junction with a countdown timer thousands of times. It’s not a problem.
Is it possible a careless driver could badly judge his/her acceleration vs the distance to the car in front and its acceleration and crash? Sure…though I’ve not seen it happen. But a careless driver could do the same at a conventional stop.
I have never seen a countdown timer.
In high school I used to caddy on weekends (remember caddies?) on a golf course beside a “Y” in one of the busy rail lines. Huge long freight trains stop and would wait for the switch to be thrown. When the train stopped, you could hear “bang-bang-bang-bang…” working its way down the train to the front. Then the train would strt again and the same loud noise would work its way back along the train to the end as each car started up again.
Traffic lights also convey to pedestrians and cyclists when they can cross an intersection more or less safely, and they don’t broadcast their speed, speed trend, direction, etc.
Perhaps the problem is mixing of time, distance and speed.
Usually one wants to define the safe distance between vehicles in time. Whatever speed you are travelling at, your reaction time plus vehicle braking performance defines the time interval you should be behind the vehicle in front. At zero speed this reduces to zero distance. At any other speed you get a non-zero time that stretches out with speed. This allows the column to reach a steady state of the same speed. In a pedantic reading of the OP, using this algorithm fits the requirement, everyone starts the algorithm when the lights turn green. Doesn’t quite meet the spirit. OTOH, in an ideal world, this is what we already do. The dwell time before you start moving after the car in front of you starts to move is essentially your reaction time - the time it takes to transition from accelerating to stopping. Even in a totally automatic system this time is never zero.
So one could meet the OP with a restatement that said, you can go as soon as the light goes green, but in doing so you must meet the constraint of never being closer to the car in front than your reaction time plus vehicle stopping distance defines. If cars violate that constraint, eventually there will be a collision.
If all the distances are fixed, the entire column can only accelerate or decelerate at the rate of the worst performing vehicle in the column. This leads to a somewhat unsatisfactory result unless you live in a new world of identical vehicle performance. You either link all the cars together into a controlled unit, braking and accelerating as a unit, or end up with shock waves propagating through the column as individual cars try to maintain distance. Not unlike rail cars taking up the slack in the couplings.
You could regard the go light as the shared control for beginning acceleration, but how you manage things after that becomes harder. Everyone could be responsible for maintaining distance, but again you get shock waves propagating, with jerky changes in motion. Or each car runs a common controlled acceleration profile and we trust that an open loop control keeps us safely away from the cars around us.
Stopping is going to be a much more fraught problem. If the traffic control system can work out what the column collective stopping distance is, it could provide the necessary command to slow and stop the column at the right point. This should include logic to divide a column, stopping only part of the column. Same problems exist for collective control of deceleration as for acceleration.