I've noticed that MH lamps , new or old seem to have a delay in striking once power is applied . MV & HPS seem to strike almost instantly .  Some of my MH lamps take 5 or more seconds to strike after turning on the power . doesn't matter if the ballast is magnetic or electronic , probe-start or pulse-start . 

 WHY ??

According to John Waymouth, metal halde lamps are hard to strike because of the presence of trace amounts of halogens in the gas. Halogens are very good at intercepting free electrons thus quenching the avalanches that pre-date discharge formation. Interestingly, a closely related effect causes halogenated plastics like PVC to burn that poorly. And the same effect also works for self-quenching in Geiger-Muller radiation counters.

But why the discharge is finally forming after a few seconds? My wild guess is that small dark partial discharges around the electrodes caused by working ignitor slowly clean-up halogen traces from the lamp volume, finally allowing the path between the electrodes to strike-through.
 

Thanks ! Makes sense . I just figured that a pulse-start lamp would light a bit quicker because of the ignitor in the circuit vs. the probe start lamp with no ignitor .

At my place of work, I have a 35W Philips CDM in G12 and most mornings, it starts without a delay. Occasionally, it will take several seconds to strike. Sometimes, however, it can take more than one minute to strike. All the ones I use at home strike within 2 seconds.

I have 1 GE multi-vapor 175w metal halide medium base lamp that cycles. It will run for 30 seconds then shut off indefinitely. I think it does that because the starter electrode was spattered or damaged. The lamps was only used for a year and it won't run on a proper ballast.

Metal halide lamps typically need a little help from outer space to strike - a Cosmic Ray.  Due to the reason explained by RRK they are more difficult to ionise and it would not be practical to use a very high energy ignitor.  The surface of the earth is continually bombarded by Cosmic rays, to which most materials are transparent.  You just have to wait long enough for one to intercept the arc tube, which provides the necessary extra energy to help the ignitor strike the lamp.

This theory was proven some decades ago by plotting the average time delay to strike vs the arc tube cross-sectional area.  Smaller arc tubes take longer to strike because the probability of them intercepting a cosmic ray is smaller.  This became a significant problem when low wattage metal halide lamps were developed.  The early samples would often wait for up to a minute or two before striking.  The solution was to borrow a trick used in fluorescent glowbottles which suffer a similar situation : to dose them with a tiny amount of radioactive material.  Its natural decay provides high energy particles with greater frequency than the mean cosmic ray interception.  Typically Krypton-85 is used at a low dosage of about 0.5-1.5mBq/litre of gas, the higher concentrations being used for the smaller lamps.  Its decay particles do not have sufficient energy to escape from the lamp so it is a very safe solution.  But about a decade ago the USA clamped down on the use of all radioactive materials in lamps, and manufacturers could no longer use that trick.  The result is that the delay times once again appeared - or in some cases a UV glowbottle was applied to provide high energy photons that help to limit the ignition delay.

At the risk of inducing some radiophobia to MH lamps collectors - I once tested a box (may be about 50 pieces) of MH lamps, mainly 70-150W types, with a large and sensitive Geiger-Muller tube. There was probably a *little* radiation increase nearby, but barely detectable, may be about 2X background.

By the way, I have read that MH lamps electrodes are made with thoriated tungsten, as the oxide activation does not survive halide atmosphere. If so, why the weak radiation of Th does not work alone to help MH lamp ignition?

Kr85 decays mainly with beta emission of average 251 keV, max 687. Not sure if an electron at such energy can exit quartz glass, but that can generate some notable x-ray bremsstrahlung. Though, quantities of Kr85 are minor to cause any concern. BTW, half-life of Kr85 is just 11 years, that could mean some really old MH lamps could have troubles starting!

Probe start metal halides (especially m57 175w reduced-envelope) Have quite a delay in striking an arc. I never experienced this issue with pulse-start quartz or CMH though. And I saw that some of my newer metal halide lamps say on the packaging: "Contains Thorium" Is thorium present in the older metal halide lamps or just the newer ones?

A lot of this ties into the 50w Quartz Metal Halide Wallpacks I use in my garage... the types of lamps I bought. My Ceramic Metal Halide Philips 50w
lamp takes the longest to strike out of all of them, and from what I see, that's the smallest arc tube. My GE QMH lamp definitely contains that glow
bottle you speak of @James. That lamp not only warms up the fastest out of all of them... it's hot restrike time is much, much shorter. I hate
hot restriking my magnetic Metal Halide lamps, but I definitely plan on doing it to show this on video. As soon as I flip off and on the switch,
that GE lamp is already visibly attempting to hot restrike from the work of the ignitor.
I'm not sure what starting method is used for my 50w Wallpacks, since I don't know as much about Metal Halide magnetic ballasted fixtures as I do fluorescent.

 Thanks for the replies everyone !  Never in my wildest dreams would I have guessed that cosmic rays had anything to do with it ! 

From experience, I find that Philips Allstart CDM lamps start almost instantly, compared to standard probe start, and sometimes even the pulse start ceramic lamps...
Also interesting on the need for cosmic rays...I've been looking at some earth sciences sites, and noticed that earth's protective magnetic field is in decline...not good for humans , but good for MH lamps making them easier to start... :

"..I've been looking at some earth sciences sites, and noticed that earth's protective magnetic field is in decline..."

I don§t think that makes that much of a difference here. The magnetosphere protects mainly the atmosphere to not get blown away, but it is affecting only electrically charged particles (mainly ions are the important here - heavy enough to have enough momentum to push the air to the outer space), which tend to be too big to pass through the quartz anyway.

And the magnetosphere is fluctuating the whole history of Earth, so I would not be worried that much. Yes, the fluctuation may have some local (in time) impact, but it will swing back again... It is also responding (or better to say the resulting magnetic field around the Earth) to the solar activity (quite a few powerful activity pattern cycles are known, my guess even many more still unknown), which is varying also.