http://www.youtube.com/watch?v=hNz7Hyz1vSs
At the copy room of the storage of Carmel hospital.
One of them are a new T8 operated with a preheat magnetic ballast with integral electronic starter (Electronic Start from Shaingot), and which aren't related to the topic of the video. The reason why its takes a long time period to restart is perhaps a defect in the ignitor.
Anyway until know we all thought that the >30 years old ballasts that drove the two T12s in this fixture in the copy room near its door (Which one of its with its ballast had replaced by T8), were air thermal starter ballasts.
However, today (13.6.2010) the electrician of Carmel hospital surprised me by say me that they were actualy RAPID START magnetic ballasts.
Look in that video at the remained T12 which still operated in one of these ballasts (The lamp that glows dimly after turning back on is the T12 lamp) and check this out: Is this behavior is corrent for a rapid start magnetic ballast (At least the european ones).
Maybe the european rapidstarts behaves different then the US ones.

I've had lights with rapid start ballasts that when turned off & back on after they've been running (long enough for the ballast to have become hot), sometimes one lamp will glow dimly for a up to minute or so then they'll both come on. (and sometimes just the lamps ends will glow very dim)

Regarding to the new T8 lamp and its preheat electronic starter magnetic ballast in this fixture: The reason for the long time of the T8 to restart (About 7 mins) is not because of any failed ignitions. This is the programmed preheating time that programmed into the starter!!! What a piece of garbage.
Have anyone heard of an electronic starter or a preheat electronic ballast that takes so long to preheat the electrodes before ignition!?!?

Regarding to the new T8 lamp and its preheat electronic starter magnetic ballast in this fixture: The reason for the long time of the T8 to restart (About 7 mins) is not because of any failed ignitions. This is the programmed preheating time that programmed into the starter!!! What a piece of garbage.
Have anyone heard of an electronic starter or a preheat electronic ballast that takes so long to preheat the electrodes before ignition!?!?

It is 7 minutes or 7 seconds?

Recommended preheat time for preheat current equivalent to rated arc current is 3..5 seconds. And as most 230V series choke ballasts do not increase significantly the current in preheating mode (lamp nearly shorted), 7 seconds would be near the optimal time for the "ideal" starter.
I know glowbottles do not preheat as long (the actual lamp preheat time is less then 0.5s, the main waiting time is for glowbottle starter to heat up and switch close, so actually start the real preheating), the consequence is then shorter cycle life.

@Medved: I made a mistake. The electronic starter of the ballast of the T8 takes 7 secs to preheat the electrodes, before ignition.
But there is a ways to short the preheating time and still ensure long life of the lamp in electronic starters and programmed start electronic ballast. Most of them preheats the electrodes for 2-3 secs.
Also Medved: why the rapidstart ballast that operate the T12 in this fixture behave like this while restarting and not like the known american rapidstart (Lighting is similar to a slow process of instant start on an electronic ballast [50% brightness during ignition, full brightness after])(At the first turn on it behaves like in this video, minus the dim glow waiting period).

Preheat time: Obviously if you use higher current, the preheat time would be shorter, only with too high current you only might run into troubles with overheating them. Preheat current of 2*Inom leading to 1sec is typical "choice" in lot of many programed-start ballasts as quite good compromise (with simple current mode preheat) for user convenience and lamp life with the assumption of manufacturing tolerances. The goal is to heat up electrodes to temperature, where (the "literature say") filament resistance is 4times the resistance in their cold state, closer you reach, longer the cyclic life. This ultimate goal you might reach by (as best) some feedback system:
- Or some active one (theoretically would work as best, but never seen it in real life), where the controller monitor the filament resistance and go to the "ignition" mode when it reach the ultimate "4*Rcold" value
- Passive one - supply the filament by the voltage source, so the heating power is reduced automatically by the increasing filament resistance by itself. This method is used in all "infinite start cycles" rated ballasts and CFL's i've seen. The supply for the filament originate from auxiliary "secondaries" on the ballasting choke. The disadvantage is, then the filament can not anymore serve as EOL protection, so the circuity need an active EOL protection, but this is usually of no extra cost, as the programmed start functionality require quite complex controller anyway, so the protection could be build-in easilly.
By the way this voltage mode heating is the "concept of choice" even in most US 60Hz RS magnetic ballasts, where it allow to operate electrodes on the correct temperature regardless of the real arc current, so allow the ballast factor to go virtually from zero to about 1.5 (arc current is split into two parts, each flow trough one side of the filament, so no part is overloaded) without

RS circuits "2-stage" starting:
This behavior is undesired consequence of the required margin for reliable starting at worse conditions (cold, humidity, thin krypton tubes,...) - the OCV is set quite high. It is so high, as at normal temperatures and dry environment the tube start in cold-cathode glow-discharge even before filaments heat up. The "golden" combination would be, then the discharge does not ignite till the filament reach required temperature. So the lamp would start with some delay at once to full power.
The consequence of this behavior is shorter cycle life.

Other reason (mainly on capacitive ballasted "HPF" types) would be the temporary lamp rectification: Each electrode is a bit different, so reach the emission temperature in different time (mainly the real temperature emission border differ here - depend on actual wear,...), so the lamp ignite for current in one direction. This charge the ballasting capacitor by DC offset, what increase the voltage of the "other" polarity, so forces some current even in cold-cathode mode. The result is, then the lamp is (upon power ON) at first dark (no discharge yet at all), then glow dimly (rectification mode; the AC voltage component is high, so the current trough the ballast is low, so low brightness) and then go to full power (both electrodes are hot enough)