The light diffusion coating of good quality integrating spheres is actually surprisingly expensive.  It is typically made from barium sulphate with a reflectivity set to 80% - this material being used on account of its relatively flat spectral reflectivity as a function of both wavelength and wall temperature, and to ensure the necessary number of internal reflections for the integration principle to work correctly with light sources having different luminous intensity distributions.  Without that, it would be necessary to calibrate the sphere differently for different lamp types and power loadings (which is sometimes still done where greater precision is required).  At the last time I had to order some of this paint, its price was a little under €2000 per litre - as such, for many years most of the lamp companies used to manufacture this paint themselves in their own chemical labs.  I had to make it a couple of times during training in my earlier career, quite an interesting process!

If the sphere is large enough, a lamp explosion is usually not too problematic because only a very small percentage of the inner surface is damaged.  More of a problem are the shards of broken glass which have to be cleaned up each time.  The barium sulphate paint has very poor adhesion, and can be brushed away with light contact.  It is out of the question to clean the inside of the sphere with any kind of brush, so the pieces had to be picked up by hand or vacuumed away.

Re-calibration is pretty simple - just install the reference lamp, give it 30 minutes to stabilise, and then scan the reference spectrum again.  This is done every few days or weekly anyway, to monitor the stability of the sphere as ambient temperature and humidity vary, but then using a checking lamp rather than the actual calibration master.  Since air conditioning has become the norm in recent years, sometimes the spheres can run for months at at time without need for re-calibration.

@James
Thank you for all of that information! I had no idea that this would be a regulated thing with requirements.

Another alternative is to stamp them on the metal ends of the tube, or metal base of the bulb. I have seen laser etched labels on lamp bases as well. Well hopefully modern etches hold up well.

The effect is extremely significant.  The luminous efficacy of an HPS lamp is determined in great measure by the amalgam temperature, which is typically set to deliver a D-line spacing of about 100 angstroms (or 120Å for the Gro-Lux types).  Higher sodium vapour pressure results in more light being wasted in the infrared, and decreased efficacy.  To measure this effect in the lab while holding all other parameters constant, we used to take simple MR16 halogen lamps and shine them onto the arc tube end where the amalgam is located, to cause modest extra heating.  Even a change of 10-20°C can drastically affect efficacy.  Hence the importance of keeping lamp dimensions and the amalgam dose weight and position absolutely consistent to avoid lamp-to-lamp variation.

In the early days when this was being investigated, and before MR16 lamps had become popular, manufacturers used to wind a small tungsten filament around the arc tube ends and provide the current to those via an extra seal in the outer bulb.  Very little heating was required to change the properties, and enable the optimum temperature to be set and recorded by pyrometry, and then tweak the lamp design until it self-heated to that point.  I do still have an example of an old 1960s HPS lamp with these end heaters.

A similar effect can of course be seen as the lamp ages - one of the primary mechanisms for voltage rise in HPS lamps (which limits their life) is the rate of end blackening around the electrodes. 

For your purposes, these effects may be small enough to live with, but it would not be a good idea for a commercial product.  Even for standard HPS lamps, the lamp and luminaire standards require that when an HPS lamp is installed in a reflector, the light reflection pattern has to be optimised such that as little light as possible is reflected back onto the end seals, and that the voltage rise of the lamp in the reflector should not exceed 5V.  Otherwise the reduction of lamp life in the fixture would be much too significant.

 Your biggest issue would be to find someone close to you as shipping would be rediculously EXPENSIVE !!!

As far as I know lanthanated tungsten was not used in HID electrodes.  As you say the efficacy is slightly lower, but we also found that we could see traces of lanthanum in the lamp spectrum.  Most lamps stayed with 2-3% thorium oxide right to the end.

What was used very widely was lanthanated molybdenum in wire and foils for HID lamps.  This showed significantly improved corrosion resistance and recrystallisation temperature, which extended the life of metal halide lamps.  It could not be used everywhere though.  I remember a big disaster when we changed to MoLa wire for the ignition antennas of HPS lamps.  It was found after some years that it suffered greater photoemission than plain Mo wire, and caused accelerated sodium loss and hence reduced life.  It was necessary to go back to pure Mo wire again.

In fact, the reason they were called etches is really because the ink was physically etched into the glass surface.  Unfortunately though the process is quite critical, and very often not done well.

The ink itself is a highly complex material uniquely prepared by the lamp industry, via an elegant chemical process that takes about 2 weeks to complete.  It basically involves dissolving the oxides of silver, lead and sometimes also copper into an organic syrup, laced with boric acid, which has the most wonderful smell you never forget after you have been in any lamp plant.  The ink is applied to the glass via conventional rubber stamping methods, and then burned in with a gas flame.  That is the critical part.  The temperature has to be high enough that the glass becomes soft, and triggers a reaction in which the lead borate chemically etches its surface, and draws the silver and/or copper ions below the surface of the glass.  The problem is that slightly excessive heating will cause the glass to deform.  It is very difficult to find the right balance, because there can be variations of one or two tenths of a millimetre in glass wall thicknesses from one bulb to the next, and the thinner bulbs are prone to deformation.  To avoid that, most setters of the Sealex machines err on the side of caution, and tend not to set the etching fires quite hot enough.  The result is that the etching reaction does not proceed fully, and the ink sometimes be rubbed away.

Many years ago this reached such a problem that too many lamps were losing their etches, which was considered a safety issue because the customer might not be able to read the wattage and know if they were installing a lamp within the the power rating of their fixtures.  The various standardising bodies such as the IEC and ANSI then developed a mandatory test for the robustness of the etch, which was taken into the IEC 61 standard for incandescent lamps and to this day has been copied over to most other lamp standards.  In the factory we have to take a soft cloth, soak it with hexane (a solvent for the ink syrup), and rub the etch softly for 15 seconds.  After testing it must still be visible.  I do think that since the 1970s most lamps probably meet this requirement, but of course we all know several exceptions where it has not been etched firmly enough.

GE lamps were another matter.  Some time ago they tried to shift to different ink types as a way of getting rid of the lead content, for employee safety reasons in handling the lead borate marking paste.  They developed a black air-drying ink based on copper oxide which did not have to be burned into the glass.  Some years later Philips also developed an air-drying ink with a golden appearance, which was rather better.  And Sylvania developed red colour inks which were extremely robust, but contained cadmium borate which also had to be dropped after a few years due to the environmental restrictions.

The reason internal etches was dropped was quite simple, because it was found that they can cause chemical contamination of the lamp atmosphere and lead to reduced life.

The problem with the glass frit decals is the same as the etch materials - they require high temperature firing.  That is fine for heavy-walled borosilicate lab glassware, but very difficult to control for 0.5mm thick soda-lime glass bulbs and tubes.  Also the speed of applying the decals is too slow and not really compatible with high speed automated production. 

@Multisubject — do you want a NORMAL MH400?

My biggest peeve is bad lighting. Ie, half-assed, poorly-maintained lighting installs. I have somewhat of a "perfectionist" nature, so if something like a chandelier has 16 different kinds of bulbs and three color temperatures, I will not be able to focus on anything else in the room, no matter how "nice" it may be.

Also flickering LEDs are up there. Recent-manufacture lamps from the likes of K-Lite (branded Philips) have a tendency to flicker horribly for no particular reason. They are straight junk in every application, and not worth more than $1/lamp, "unique" attributes aside (ie, SceneSwitch).

I also can't stand missouts. Like you went out of your way to replace every light in town. So WHY is there a light smack in the middle of it all, that is probably EOL, that is still in place??? My only exception to this rule is when a light left lit is something like a vintage MV lamp (ie, a local town converted entirely to LED, and they missed a Westinghouse OV-15 that's been reliably lighting dusk to dawn for the past 25 or so years).

There are more, but that's all I can think of so far.

 I've seen missouts here which were caused by blocked access to said fixture during the changeout process .

@James — Thank you! Interesting to see their company philosophy compared to a company like GE who built their factories in somewhat larger towns, and thus built them to a larger scale. In my opinion, from the 1970s onward, I would say GTE had the better finished product. Lamps made in Ravenna, OH, were DECENT, but had nothing on Sylvania's Manchester, NH-made lamps. And I assume different parts of each lamp were made in different places? And did Sylvania make their bases in house, or source from someone like GE?

@Ash: Isn't the CRI of SOX lamps Ra-44?

I had a Ledvance A60 13W 6500K snow cone LED lamp that lasted exactly 1 year of almost 24/7 operation at base-up burning: https://www.lighting-gallery.net/gallery/displayimage.php?pos=-252219