@Medved -- Good to know! Seems logical that they are engineered to last no longer than the lifetime of the car. I simply find it bizarre that OEMs are sourcing lower-quality assemblies which fail in these bizarre ways... on newer vehicles nonetheless. I think that Corolla I mentioned is still under warranty, so I assume it would be replaced if brought to a Toyota dealer... I find it interesting that it's also an anomalous process, and only few cars are actually plagued with the issue, and not even both lights... this variation and seeming "randomness" is what I find most intriguing. So bizarre to me that they do not fail in pairs as one would expect.

Generally there are two types of semiparallel ignitors, they differ in the ballast winding configuration they require:
One uses the R-C-SIDAC configuration (or equivalent function using more complex circuit allowing additional features like timed EOL shut down or so). These need tap on the ballast winding of about 5% from the lamp end. So far have seen this only in literature, not real commercial type, but I did not researched them that much. My assumption was, adding the separated pulse transformer and making a superimposed ignitor out of this is the more common way.

Then there are circuits based on the semiresonant concept , which are used mainly because of their ability to drive larger capacitance, so longer cable towards the lamp. These use ballast with tap at the 20% from the line input. An example is Philips SN57, but there are multiple implementations, usually varying in the "controller" part.

Fact is, the automotive headlights are driving the LEDs way harder than it is common with general lighting. Paertially because auto headlights needprecise beam control which needs high intensity packaged to a small area, partly because car headlights use to be designed by far shorter lifetime than general lighting, because simply the cars do not run that many hours over their runtime. Normal passenger car lifetime of 300000km/200000mi means about 5k hours running, headlights are on for barely 1/3 of that, so lifetime of 2k hours means the headlight will last the whole car life, which means less than 1/10 of typical lifetime requirement vs general lighting. And the headlights are really engineered to last just that, to get the required performance at the lowest cost possible, so some degradation effects may already show up, mainly when they do not constitute headlight failure.

And the hard drive may mean different degradation effects than usual for general lighting become dominant. And then there is many years of off state parking time, when the humidity may corrode things without being drawn out by the heat of normal operation.

Not on the tipic of pizza, but rather the original topic.

Went down to a pub at the town that I'm in at the moment, had a Chicken Parmi with a side of hot chips ('fries' for Americans).

Cici's Pizza is still a thing in some locations, although I haven't seen them in a very long time. The one we had near me in Fairfield commons many years ago
was gross though, unfortunately.

Remember when CiCi’s was a ‘thing? We used to get ads for them in CA however we either never had or used to have locations here. I always thought their cinnamon roll looking pizza was interesting.

Both lamps emit their IR as somwhere between a "line source" and "point source". Which one of them closer describes your case depends on how close the lamps are to each other. If they are close, let's assume it is line source

Line source : Irradiation is reciprocal of the distance from the source

Point source : Irradiation is reciprocal of the distance from the source squared

Assuming an 1.5KW linear Halogen lamp within 1" from the outer glass of a 400W SON, which itself is a ~T17 tube containing a 8mm thick arctube. The arctube is about 1/2 of the length of the Halogen. (I am thinking here in terms of lamps common here, your exact lamps and sizes may vary)

So we have 2 possible paths of heat transfer by IR :

 - From halogen lamp to arctube (for wavelengths for which the SON glass is transparent)

 - From halogen lamp to SON outer glass, then separately from it to arctube (for wavelengths for which the SON glass is opaque)

We can assume (and it is close enough to reality) that the entire 1.5kW output of the Halogen is in the IR. Though i dont know how it then divides between IR for which the SON glass is transparent vs opaque



Transparent glass :

Arctube's part of the projection from the halogen :

8/(2∙50.8∙3.14)/2 = 0.013
Where :
8 - Arctube diameter
50.8 - Radius at which arctube is from the halogen
/2 - The arctube is 1/2 the length of the halogen

0.013∙1500 = 19.5



Opaque glass :

SON outer's part of the projection from the halogen :
60/360/2 = 0.083
Where :
60 - For T17 SON and a Halogen at 1" from it, the SON obstructs 60deg of the perimeter of the halogen
/2 - The SON arctube is 1/2 the length of the halogen. (The outer is more, but then not all the radiation inside will get to the arctube, so this can be still factored in here)

Arctube's part of the projection from the inside of the SON outer glass :
8/(2∙21.4)/2 = 0.093
Where :
8 - Arctube diameter
21.4 - Radius at which arctube is from a linear strip of the glass
No pi this time as the strip of glass is flat emitter, so the intensity seen by the arctube is according to cosine law
/2 - Half of the radiation is radiated to outside the SON

0.083∙0.093∙1500 = 11.6



So, the overall power gained by the arctube from the halogen by IR will be : X*19.5 + (1-X)*11.6 , with X between 0 and 1 depending on which part of the IR spectrum the SON glass is transparent to. Worst case is X=1

So the SON arctube gains extra 20W from the irradiation

The additional delta T caused by heat transfer through other means (convection heating the SON outer, and then by IR from it to the arctube) is probably insignificant

I'd say the effect is about same as adding +20W electrically to the arctube

From ignitors i have :

ELT semiparallel ignitor - based on some generic thyristor, fairly simple

Steinitz ST1C - based on BT138 thyristor + few resistors and capacitors

Eltam ES50 - based on Littelfuse K220 SIDAC, with the exact circuit from the datasheet. Probably identical to American ones. https://www.lcsc.com/datasheet/C151184.pdf

And then we have Eltam's ESPI1000, with a PIC microcontroller inside. (The power stage is about same as in the ELT)

So it varies. The complicated ones implement cut off timers and logic for few purposes :

 - Failed ignition attempts still do make some discharge in the lamp. Without them the lamp cools faster, so this might speed up hot restrike time

 - Generating high voltage continuously (and especially without a functional lamp that would clamp the pulse) stresses the insulation on the ballast for no purpose. (Tho most decently made ballasts survive it for years anyway)

 - People dont like cycling lamps, so some prefer a lamp to stay out completely if it failed. (On the other hand, any unmaintained lot with multiple cycling lamps, will still be more lit at any given time with the simple ignitors)

I know it's simple, but a big meatlovers fan over here! I also like a good supreme, but keep the pineapple and anchovies off!

Another favorite is a bacon and egg pizza, essentially a bacon and egg roll, but on dough.

Nothing like farm fresh chocolate milk. Amazing stuff.

Nah chocolate milk is actually quite good! Always reminds me of Denmark!