In my experience with taking pictures of HID lamps running off North American CWA ballasts, I have noticed that while lamps running on North American high pressure sodium, probe start metal halide, and pulse start metal halide CWA ballasts produce the same type of flash pattern, I have noticed that lamps running on North American mercury vapor CWA ballasts tend to produce a different flash pattern that has wider dark bands compated to the high pressure sodium, probe start metal halide, and pulse start metal halide CWA ballasts.


Why are the flash patterns so different between North American high pressure sodium, probe start metal halide, pulse start metal halide, and mercury vapor CWA ballasts?

This is what the flash pattern looks like when lamps are running off North American high pressure sodium, probe start metal halide, and pulse start metal halide CWA ballasts. In this case, the lamp is running off a North American probe start metal halide CWA ballast.

This last image represents the same lamp running off of a North American mercury vapor CWA ballast. Notice the wider dark bands when the lamp is running off the mercury vapor CWA ballast.

CWA ballasts have different waveform than the mains voltage.

The mercury vapor sufficed with the lowest OCV and were the most robust on the high current crest factor, so the ballasts utilized these benefits to allow them to be cheaper and less lossy. What you see is the high current crest factor (all current concentrated to just rather narrow spikes). The other lamps would either degrade too fast (MH - the electrode sputtering would be a killer) or the associated long gaps without the current would lead to unstable arc (mainly HPS, but the MH as well), so the ballasts for these need to sacrifice some extra losses and cost, in order to deliver less harsh current to the lamp.
All is about how much voltage comes from the secondary only (higher voltage tap on the primary where the bottom secondary end connects means shorter wires so lower losses, but less of the voltage could be influenced by the magneticshunt leakage), what is the exact shape of the saturating part of the shunt (sharper saturation allows better regulation with less voltage at the secondary, but yields harsher current shape to the lamp) and so on.

@ Dor:
That wasn't the question that was asked. Medved has given a very thorough answer on the subject of why there is a different flash pattern between lamps running on the same type of ballast.

The mercury vapor sufficed with the lowest OCV and were the most robust on the high current crest factor, so the ballasts utilized these benefits to allow them to be cheaper and less lossy. What you see is the high current crest factor (all current concentrated to just rather narrow spikes). The other lamps would either degrade too fast (MH - the electrode sputtering would be a killer) or the associated long gaps without the current would lead to unstable arc (mainly HPS, but the MH as well), so the ballasts for these need to sacrifice some extra losses and cost, in order to deliver less harsh current to the lamp.
All is about how much voltage comes from the secondary only (higher voltage tap on the primary where the bottom secondary end connects means shorter wires so lower losses, but less of the voltage could be influenced by the magneticshunt leakage), what is the exact shape of the saturating part of the shunt (sharper saturation allows better regulation with less voltage at the secondary, but yields harsher current shape to the lamp) and so on.

Actually, in North America, high pressure sodium CWA ballasts have the lowest OCV at about 180v, the mercury vapor CWA ballasts have an OCV between 198v and 350v, the pulse start metal halide CWA ballasts have an OCV between 265v and 300v, and the probe start metal halide CWA ballasts have an OCV between 280v and 330v for lamps with a wattage between 150w-400w. However, for the higher wattage CWA ballasts for 700w-1500w lamps, the OCV is between 370v and 480v. I believe that in this case, the OCV seems to have no effect on the flash patterns.

With the OCV I actually meant the ratio betweenthe OCV and the lamp arc voltage. So because the high power MV have high arc voltage, the OCV then needs to be higher. But indeed, it is definitely not all to OCV (even when meant that way) alone.
The thing is, you may get certain OCV (assume you want 220V) by either starting on the 120V top of the primary and add 100V on the secondary. Then only the 100V part is going to be controlled by the shunt. Or you may start from e.g. 80V tap on the primary and add 140V on the secondary. That way you get the same 220V OCV, but the whole 140V section is behind the shunt, so responding to the current. So these two example ballasts will have the same OCV (220V), even have the shunt tuned for the same leakage inductance (so the ballast running with the same capacitor), same current when the shunt saturates, but will respond to the current in a different way (assuming all other construction the same, the second will likely be softer, so lower crest factor and a bit worse regulation). And will exhibit different losses (the second will have longer secondary winding wire, plus the primary wire currents will be higher as well).

Or the difference could be just in how the magnetic shunt is shaped: Two ballasts, exactly the same windings (same wire, the secondary is connected to the same primary tap, really exactly the same) and main core, using the same capacitor, operating the same arc voltage and same lamp current. The only difference is in the magnetic shunt design: One has the shunt made as a rather sharp cur rectangular block, the second one has its edges rounded (both are set so they yield the same rms current). That means at the point it reaches saturation, with the first one the whole shunt will saturate at once, so the leakage inductance will go down very steeply. This results in sharp response to the current (so will have good mains voltage variation suppression even with low voltage from the secondary, so low losses) but will exhibit large current spikes (once saturated, the inductance limitting those spikes gets really low). With the roundy one, only the tip will saturate first, but other material is still not much further away. As you increase the flux, gradually bigger part saturates, so the inductance reduces gradually. That means even at current peaks there is significant inductance available to smooth them out, so you get low current crest factor. But because the inductance changes only gradually, the mains voltage suppression is not that great. So often it needs more of the secondary to be controlled by the shunt, so yields a ballast with higher losses (or heavier and more expensive).

Bottom line all these things are about a compromise between the mains voltage tolerance suppression, current crest factor, ballast losses, cost and mass. Plus often to "stick" to the use of standardized components (core plates, bobbins, even the complete coils,...) to lower the cost of part logistics during manufacture.