http://picasaweb.google.com/110595932802324922410/Lighting?authkey=Gv1sRgCI6D9bav-ezxcQ#5456725038699463266
Which type of ignition mothed the lamp ignited? (To me this look like a rapidstart ignition).

Instant start. It first glow shortly in cold-cathode mode (a bit dimmer, with bluish end-glow) and then (after ~0.5s) it shift to hot cathode discharge, lamp appear to start "from the middle".

RS would be the ~0.1..0.2s delay and then directly hot cathode mode ( "single step" runup - no intermediate dimmer phase, no blue end glow, the lamp appear as starting "from the ends").

RS (heating filaments, where the lamp voltage is not enough to ignite cold, but just enough for warmed up cathode lamp) is very rare on electronic ballast, as these or use directly programmed start (0.5..2s delay with very low lamp voltage unable to ignite even with hot cathodes, with yellow end-glow preheat and then the arc strike directly to the hot-cathode mode) or instant start (very high starting voltage, able to ignite even cold lamp and able to run it in cold cathode mode)

Medved: Fluorescent lamps that lit on rapid start magnetic ballast usually have such behavior like this lamps. But the dim phase of the ignition before the lamp is finally ignited is more longer then this lamp. Also, instant start CFLs usualy ignited and lit steadly from the moment the switch is flicked (like the incandescnets and LEDs when lit) (Instant lit without any early states before the ignition itself [As they accured in this lamp]).

The intermediate state is a result of the cold cathode discharge, where large cathode drop cause the overall lamp voltage to be larger then the ballast is designed for, so the current is smaller, yielding lower dissipation in the anode column. Now it is the question, what current is delivered to the lamp in this phase and how long it take, while the second is strongly influenced by the first (larger current in the cold cathode mode mean cathodes are heated by larger power, so warmup more quickly, while at the same time the brightness difference might not be as visible).
On simple selfoscillating electronic ballasts the high lamp voltage mean, then the filament current is large as well, what help to heat up electrodes, so speed up the cold cathode phase as well.
But some designs do limit the striking voltage (inherently, or on purpose as a ballast fire protection measure), what lower the cathode dissipation, so prolong the duration of the cold cathode state.

So if i'm right, the CFLs and the linear fluorescents operated with a chinese low cost ballasts that lit instantly like the incandescents, halogens and the LEDs do the intermediate state of the cold cathode discharge in less then 0.001 sec before they finally reach hot cathode state, so this cold cathode state of the ignition in them is invisible to the naked eyes and the lamp looks like its lit instantly at hot cathode state.

Better to say: Even in the cold cathode mode, the current, so lamp brightness, is very close to the normal operating value, so there is no perceivable brightness difference in the later cold->hot cathode mode transition.
But one thing you might observe on some lamps: Bright glow in the area "behind cathodes" disappear after transition to hot-cathode mode. But don't be confused: On some simple ballasts this glow persist as a result of cathode overload by high crest factor current waveform (some lead type low frequency ballasts - like series capacitor in some night lights)

Medved: I saw this bright glow around the electrodes only in capacitive ballasted CFL nightlights and in several emergency fluorescent lanterns during turn-on. In most of my instant on CFLs and in fluorescent fixtures with instant start ballast of the same behavior i don't saw any bright glow around the electrodes when turned on. They lit instant and smooth. Only at the two F8T5 lamps fixture in the bathroom at my mother home, the lamps flickered somewhat after turning-on.
By the way, is low current cold cathode mode for instant start electronic ballasts prolong the electrodes life when comparing to high current cold cathode?

By the way, is low current cold cathode mode for instant start electronic ballasts prolong the electrodes life when comparing to high current cold cathode?

It is not as simple answer.
- The real cold cathode wear progress is proportional with amount of ions, so with multiplying current and time, provided the temperature profile is constant
- Higher current mean, higher temperature is necessary to handle it in "hot cathode" mode, so the electrode need more heat to reach such temperature, what prolong the duration of the cold-cathode mode, so increase the associated wear.
- Lower current mean more significant is amount of heat radiated away from the electrode, what prolong the duration of the cold-cathode mode, so increase the associated wear. Moreover relatively longer time is spent in the phase, where the temperature is high (so materials are weaker and easier to sputter away), but still not high enough to reach emission necessary to lower the cathode drop.

I would say there would be somewhere an optimum, but i doubt somebody really care of reaching it - instant start are designed or to be cheap (so nobody care as much for the lifetime) or for applications with long burning hours per start, where the starting related wear is not a factor, unlike the extra power dissipation for "better" ballasts.
It is ironical, then those simple, selfoscillating, instant start ballasts are those most energy efficient. There are two reasons: First, they do not heat up filaments (about 1W for F36T8), seconds, there is no controller needing it's supply (~1W for most controllers) and thirds, the selfoscillating concept allow the use of bipolar transistors, what have way lower "equivalent Ron", so associated losses, while their slowness is not an issue, as these ballast work in "soft switching" mode, entirely eliminating (otherwise high for bipolars) switching losses, so MOSes loose the game... Issue is, then those more advanced IC's are not able to drive bipolar transistors (or it would require more supply power for the controller), so they are stuck with otherwise more lossy FET's.