At the Crossings shopping center, some of the storefronts have those round recessed metal halide fixtures. I’ve noticed some don’t work, some produce the typical white MH color, but others are pinkish tint.

Do work metal halide bulbs ever produce this color? Mostly I have seen worn MH bulbs making a green color.

Also I’m guessing some of these bulbs could even be the original ones from 2007 when the center was built, but IDK.
It doesn’t appear they maintain the canopy lights well which is odd given they’ve replaced the parking lot lighting with LED. One of the fixtures is even having out of the ceiling!

Yes, it depends on the brand and what halide salts they chose.

That is a classic sign of a very old but also very well made sodium-scandium metal halide lamp.  In cheap or basic lamp types, sodium loss is one of the primary failure mechanisms - the removal of its orange wavelengths results in a greenish shift.  However in NaSc lamps scandium loss reactions are also significant.  Often they are overdosed with an excess of scandium but if not, or if the sodium loss reactions have been minimised by excellent design and materials, they will turn pink as they age.  If left to burn longer they can even become orange and look like an HPS lamp.  There is a photo of such a lamp somewhere on this forum but I cannot find it now.

That leads to an interesting question the answer to which I always wanted to know. If I understand the physics right, then if a kind of halide salt, say NaI, is kept partially in the molten state, sodium vapor pressure in the lamp depends only on the salt pool temperature (saturated vapor) and does not depend on the exact physical amount of NaI salt introduced into the lamp, right? So, if we make a hypothetical MH lamp of DyI3 (boiling point 1320C @1bar) and NaI (BP 1304C @1bar) then Na / Dy ratio is pretty much fixed and depends only on the burner cold spot temperature, yes? Then how halides ratio is adjusted in real lamps? One can put a small amount of the salt that would evaporate completely, but this is expected to be unstable? 

@James
I think I found the image-
https://www.lighting-gallery.net/gallery/displayimage.php?album=search&cat=0&pos=26&pid=214929
It is completely orange.

That is how saturated vapor lamps work. The key and also the weakest point is to maintain the pool temperature at that exact level. With regular fluorescents it is rather simple - because of the rather low power density it assumes an ambient temperature (plus some small difference), so it is just matter to compose the amalgam right.
In high pressure lamps it becomes a thing of very tight thermal balance, to big extend utilizing the temperature stabilizing property of the radiated heat to be proportional to the 4'th power of the temperature. And the biggest problem is how to deal with component aging and its impact on things like radiation absorbttion coefficient (steering the heat radiation), absorbtion of the arc radiation (so the heat the pool spot acquires; causing the blackened tube to run hotter) and generally the real power transfers.

@LAB27:  Yes that’s the photo I was thinking of, well found!

@RRK: It’s an exceptionally good question which Medved has already answered very well.  I can only add a little more.

What determines the quantity of halide in the arc is not its boiling point but its vapour pressure.  Even some solids produce vapours (think of solids we can smell).  The starting point for metal halide lamp design is an atlas  (produced by Venture’s subsidiary APL with additions in dozens more scientific papers) of how the vapour pressures of the salts of interest change vs temperature and physical pressure.  Since the 1980s-90s that has become pretty well known for each of the individual halides.

When we then prepare binary mixtures, the vapour pressure of one component is influenced by the other.  Exactly the same principles apply as in alloying of metals, in which the melting point of the alloy is typically lower than either of its components, and changes with composition.  A 2D phase diagram then has to be made to illustrate how MP changes vs composition.  The same can be made for vapour pressure.  Ternary and quaternary mixtures are the norm in most metal halide lamps so we need 3D and 4D phase diagrams - and hundreds of metal halide lamp engineers around the world have served almost their entire careers on that line of research.  Since the 1980s it has been done by computer models built by each of the lamp companies but usually always shared between each other.  The results can then be plugged into the lamp design modelling software which is usually more confidential.

Conclusion : adding more of one component to a big salt pool in a lamp can still change the vapour pressure because it changes the chemical composition of the melt, and the weak intermolecular forces that govern vapour pressures.

A second explanation is that the temperature gradients in metal halide lamps are often extremely sharp.  At the cold spot where the halides reside, gradients well over a 100K/mm are not unusual.  So increasing the volume of salt changes its temperature and vapour pressure.  Additionally if you look at a burning metal halide arc tube (best by using a projection lens to image it on a screen) you see that the halide is not usually restricted to a well-defined pool with sharp edges.  Due to surface tension and energy influences with quartz/PCA a kind of mist-like deposit may creep as much as half way up the wall.  The distance is influenced by the total quantity of salt and since such areas are very hot (but only contain a tiny amount of halide) this also contributes to the quantity of vapour entering the plasma.

Moreover a bigger pool also absorbs more radiation from the arc and is further heated beyond what might be the coldest temperature at an infinitely small spot.

Probably there are other parameters as well but since I did not work so much in that area I only became aware of the general principles.  I will definitely ask some of my remaining lamp chemistry colleagues at work for further experience!