I think some of the Japanese lampakers used it.  For sure the Toshiba 20W ceramic metal halide of around 1997 used submicron alumina.  Also the very first briefly made SLI CMI-T lamps of 1997.  But I think no other companies! 

1) The general failure mechanism is glass cracking.

2) Borate glass is known as a so-called “short glass”, i.e. its working temperature range is very limited.  When heating by flame, it transforms from solid to a low viscosity liquid very quickly, kind of like how ice suddenly melts into liquid water.  I understand this is due to reduced ionic bonding presence in the glass when hot.  It is predominately held together by Van der Waal’s forces when hot, which are weak, so it has low viscosity.  This makes is extremely difficult to work by traditional processes for making glass tubing, on Danner or Vello machines.  In the melting pot it is fully liquid, and when flowing onto the tube forming mandrel it is either so liquid that it drips to the ground, or so solid that it can no longer be drawn off the mandrel.

Philips developed a new technique of extruding borate glass rods in the 1960s.  After about 20 years of intensive efforts, they finally managed to extrude borate glass tubes.  From that moment onwards, a pure borate sleeving became feasible for SOX lamps, and in the late 1980s the so-called SR (sodium-resistant) seal was developed.

Magnesia is used rather than alumina due to its lower sintering temperature and hence cost.  In the later 1990s I did develop alumina beads for SOX lamps at GE, based on using scrap alumina powder that wasn’t quite good enough for our Stellox arc tube production.  It worked rather well.  But to scale up for production would have required expensive tooling.  The financial saving was too small, so never proceeded and we remaimed with magnesia beads.

3) For the hard glass sodium arc tubes, we BTH-Mazda developed the glass type C42.  It had excellent sodium resistance and also a pretty good working temperature range.  It was possible to draw into single-ply tubing.  So the seals were also sodium resistant, and no need for a ceramic insulator tube.

@James
Ah-ha! So they did try it! Very cool, I would love to see a comparison. Unfortunate that they didn't see commercial use, but there is of course a good reason for that. Thanks for explaining!

@James, For some odd reason, I have been seeing that the 1992-1993 National Catalog claims that their energy saving mercury vapor lamps have the same lumens as the lamps they intend on replacing.

For example, I see that both the National 250W HF250X and 235W HF250X/235 lamps are rated for 12,700 initial lumens.

Additionally, I also see that the 300W HF300X and 280W HF300X/280 lamps are both rated at 15,800 lumens.

The 400W HF400X and 375W HF400X/375 lamps are both rated for an initial lumen output of 22,000 lumens.

The 700W HF700X and 660W HF700X/660 lamps are both rated for an initial lumen output of 41,000 lumens.

Lastly, the 1000W HF1000X and 940W HF1000X/940 lamps are both rated for an initial lumen output of 59,500 lumens.

I wonder why National decided to rate their energy saving mercury vapor lamps to have the same initial lumen output as the full power lamps they intend to replace. The information can be found on pages 1088 to 1095 of the 1992-1993 National Facilities and Outdoor Lighting Catalog.

See here:

https://esctlg.panasonic.biz/iportal/CatalogViewInterfaceStartUpAction.do?method=startUp&mode=PAGE&catalogCategoryId=&catalogId=792140000&pageGroupId=1&volumeID=PEWJ0001&keyword=&categoryID=&sortKey=&sortOrder=&designID=retireOut_sp&designConfirmFlg=

It seems to me that this catalog contains totally false information.

It was tried also with Alumina, ending up with submicron particles.  But as RRK stated, less cost effective.  Also, the submicron alumina arc tubes are less transparent than the YAG or YbAG.  At work I am pretty sure I do still have a tray of empty arc tubes of near-transparent submicron alumina tubes, along with YAG.  When I find these I can make a photo to compare the appearance.  They were made by  Dr. Yanagitani’s process at Konoshima in Japan and he offered some prototypes when visiting my lab about 20yrs ago.  His patents and technical papers explain the technical challenges, along with similar from NGK Ceramics and Toto Ceramics.

The reality is that the so-called “Energy saver” mercury lamps are also “lumen savers”.  In fact, their efficacy is always worse than the lamps they replace, so if the power is reduced by 25%, the lumens drop by more than 25%.  Moreover, the ballast efficacy is also reduced, so the system efficacy is even worse.

Changing the arc tube fill from argon to krypton or xenon would have no significant difference.  During operation the rare gas filling is not ionised.  In HPS lamps we can improve arc tube efficacy by using higher pressures of higher molecular weight gases, because they reduce heat conduction losses to the arc tube wall.  But general lighting mercury lamps are not wall stabilised, there is much less heat loss to the wall, so the opportunity to improve efficacy by the same technique is absent. Possibly however the life could be extended.

@Lcubed3, I've thought the same. This entire living room is lit with eleven 2-lamp F30T12 strip fixtures, running F30T12 Cool White Halophosphate lamps, so... not the best in terms of CRI.
Although this still looks better in person, it does look... colorless. I know for a FACT that my friends would mention the same. There will be a little more color added as time goes on... this is still incomplete.
But the boneless couches was a big part of this idea.
Just a heads up though, to anyone who's wanting to get these... if you're needing to carry them up the stairs, be very careful.
I'm physically strong and I can tell you guys right now that they both seemed heavier while they were compressed... I remember almost falling backwards down the stairs due to the weight.
Had that happened, it would have been deadly. So yeah, be careful if this requires going upstairs. You may need a second person to help.

@LightsAreBright27, did your 9w and 11w PL-S lamps contain capacitors near the glow bottle? (Really interested in that 11w one as well, since
it appears we do not have 11w PL-S lamps here in the U.S.)
Also, when you say the EOL lamp flickered, do you mean flashed?
I would think that an EOL PL-S lamp, at least on a non EOL protected electronic ballast, would be extremely hard on the electronic ballast to
the point of ballast EOL.
... but what was the point of adding a capacitor near the glow switch bottle on PL-S lamps? I know now from you guys, @RRK and @Medved, that those
capacitors can enable use with certain electronic ballasts, but do they not help in terms of performance on the magnetic Preheat ballast?
Why add a capacitor for the starter based PL-S lamp if it's not required?

I did an experiment a while back, using a philips programmed start electronic ballast on pl-s 9w and 11w lamps. The ballast specified it can be used with those (out of 4 pins, only 2 should be connected). The lamps worked, and some even started with a starter 'ping' just like a magnetic ballast. I put an eol lamp, and it flickered as if it was on a magnetic ballast. After 10 seconds the ballast turned off (eol protection)


So an electronic ballast is able to run a standard pl-s lamp with a glow bottle starter.

In our town all carbon arc lamps were replaced around 1925 to high power incandescent lamps. Most wall brackets for these carbon arc lamps are still in situ but of course now with more modern street lights. One of them has a replica carbon arc lantern with CMH.
A carbon arc lamp I added last year into our lighting museum. It needs to be restored and many parts such as the glass holder are missing.