http://www.youtube.com/watch?v=T0YZh109sck
This seems to be a superimposed ignitor used here. However there isn't any cold cathode spark during the ignition. Why?

There always is, before electrodes heat up (~2sec after ignition).
What is here rather unusual is, then the first tiny, low energy spark always trigger the high current arc.
The reason might be rather low ignitor voltage, but very good phase timing. So the tiny spark does not jump over before the arctube really cool down to quite low temperature. But in such case the voltage drop in cold cathode mode is already below ~150.200V, so the ballast is able to immediately build up the high enough current to sustain the discharge even after current zero crossing point. The ignitor phase timing is important in order to trigger the discharge in the moment, when the mains voltage is high, but still long enough portion till zero cross remain to allow current build up in the ballast.

@Medved: This lamp flashed one time secs before i started to photograph this video. Sorry that i forgot to use the optical zoom that would make a better results. I simply can't change the optical zoom during the video photographing.
I saw some of these lamps in park Hecht before cycling (The ones with the Na-Sc and the phosphor as picture here). They reignited relatively fast, often with a successful reignition from the first try (No flashing or sparks at all during reignition, after the cooling down period, just instant light after total darkness). As a result, i always couldn't be able to make a video of them reigniting.
Perhaps these lanterns uses another type of ignitor (Not superimposed).

If they reignite always without any flashes, they likely use timed ignitors (regardless if semiparallel or superimposed, this feature is offered for both): After some time of unsuccessful igntion attempts (usually about 10 seconds) they stop and after given time (usually 5 or 15minutes; 5minutes are intended for HPS, 15minutes for MH) they resume ignition attempts.
The 10second time is designed to be longer then required for cold lamp striking, the 5 (or 15) minutes longer then the time required for lamp cool down below it's temperature limit for reignition (so it successfully strike immediately when the ignitor retry it).

Sometimes these ignitors are equipped with another timer counting cumulative time of reignition attempts and if this reach given value, they stop till the power is switched OFF. This is set so, it allow about 3..5 restarts, what is assumed as sufficient for random dips in daily power cycle (e.g. streetlights), while it effectively shut down cycling lamp.