That is indeed how i imagine it. There is (unbalance L) + C series circuit, which does not have to be in resonance but to pass the desired current for filament heating. Then the rest of the winding steps up the voltage

The phase diagram may vary somewhat between manufacturers - Eltam use a different capacitor value than UK manufacturers. Also, i have seen moderate++ success with Eltam ballasts starting 36W T8's, something which is not supposed to work according to descriptions by UK users

Would definitely like to dig up mine and provide here whats possible

Sure there may be some multiple variants of SRS circuit tuning, for the circuit in Atlas book, in staring mode the voltage on the primary winding is just 40V so apparent 'resonance' is really weak. BUT if you look at vector diagram you'll see the voltage at choke winding and at capacitor is in-phase, which should not be a case for a series resonant LC tank. So may be a better model for the circuit will be a step-up auto-transformer (fed by a capacitor current) made from primary and secondary windings unbalance?

If the winding would be assymetrical, then it would matter which winding is connected where. But as I understood, the wiring diagrams do not explicitely mention that to matter, so my guess was they are symmetrical.
Plus SRS was developed in the T12 time and T12 were happy with about 200V for ignition with warm electrodes, so you don't need any voltage boost there, even with 220V nominal mains.
And I doubt the design won't prefer some extra "poka-yoke", when there isn't anything explicite to not allow it.

only checked one of my SRS ballasts so far, a Multi-wattage Parmar 5ft-8ft 65-85W jobby

https://www.lighting-gallery.net/gallery/displayimage.php?album=lastupby&cat=0&pos=27&pid=261057&uid=476#msg889328

and as you can see from my comment in the upload above, the windings are not of the same DC resistance so I would go out on a limb and say they are asymmetrical

This damn circuit is quite hard to understand, despite being that simple. Vector diagrams in Atlas book from @LightBulbFun help a bit.

Despite a loud name, not much resonance is going on here. For a simplified variant, one can ignore filament resistance and assume the windings are the same with 100% coupling. That way, on startup, the circuit degenerates just to two filaments in series with a capacitor, powered from mains, and in working mode, (surprise!) to just a trivial lag single choke circuit with a PFC capacitor across THE MAINS (!). A classic circuit theory approach where you are allowed to mentally connect the nodes with the same potential helps with the analysis.

From Atlas diagrams, though the windings have some slight asymmetry, and sure coupling is not 100%, which will give a weak resonance kicking OCV up a little.

I find it interesting that claims are made that there is no or little resonance going on when the atlas technical booklet explicitly mentions that they work by resonance to boost voltage for lamp starting

in particular what do you say about the comment in the technical booklet about the voltage rise coming from the adding of Vc to Vs? 



also should be noted SRS ballasts for 18W SOX lamps where quite common (as they officially require about 300V to strike, so an SRS ballast was seen as a good way to generate that in a compact low-loss-ish ignitor-less setup that also has the added benefit of inherent PFC correction)


https://www.lighting-gallery.net/gallery/displayimage.php?pos=-116224


(I was always a bit annoyed how they tied the coils together into a single terminal like that, so you cant use it with a regular 2ft tube or 18W-24W PL-L or 28W 2D if you wanted to LOL)

This damn circuit is quite hard to understand, despite being that simple. Vector diagrams in Atlas book from @LightBulbFun help a bit.

Despite a loud name, not much resonance is going on here. For a simplified variant, one can ignore filament resistance and assume the windings are the same with 100% coupling. That way, on startup, the circuit degenerates just to two filaments in series with a capacitor, powered from mains, and in working mode, (surprise!) to just a trivial lag single choke circuit with a PFC capacitor across THE MAINS (!). A classic circuit theory approach where you are allowed to mentally connect the nodes with the same potential helps with the analysis.

From Atlas diagrams, though the windings have some slight asymmetry, and sure coupling is not 100%, which will give a weak resonance kicking OCV up a little.

Hi again, this isn't too helpful, but I did just acquire the bottom half of one, if you do end up scrapping the 2nd one, I'd love to get the lampholder and its stalk, again, not too helpful though.

If the winding would be assymetrical, then it would matter which winding is connected where. But as I understood, the wiring diagrams do not explicitely mention that to matter, so my guess was they are symmetrical.
Plus SRS was developed in the T12 time and T12 were happy with about 200V for ignition with warm electrodes, so you don't need any voltage boost there, even with 220V nominal mains.
And I doubt the design won't prefer some extra "poka-yoke", when there isn't anything explicite to not allow it.

Small flashtubes are typically powered by discharging an electrolytic capacitor with circa 300V voltage.  Charge circuit is some kind of current limiter, voltage doubler I suppose in US with 120V mains, or a small step-up converter for DC or battery power. Like already was said, triggering is done with a small pulse transformer, generating short spikes of a few kilovolts by discharging a small circa 0.1 uF capacitor to transformer primary via some negative resistance element. Negative resistance may be a neon bulb, SIDAC, SCR etc.

You check ~300V at the capacitor first, if present, triggering circuit shall be verified to work. HV transformer may get shorted, obviously.

I think there might be a micro fracture in one of the solder traces on the board, because in some spots it looks slightly delaminated. Weird, this ballast has only about 2000 hours on it! But it is one of the earliest HF performers they made. I know they had some problems with the early generation.

Yep, I also think a solder crack (and consequently also trace crack) is indeed quite hot candidate for a fault. It checks a lot of boxes:

Is acting up only at some temperatures
It appeared after some time
Mainly when the thing was operated outdoors, so went through wide temperature swing cycles (cold winter nights vs hot summer days or evenings) or even worse mechanical vibration (wind vibrating a pole,...)
Sounds like part early production using lead free solder (it took about the two decades to iron the reliability problems out and settle to some usable solder composition, still still there is some engineering work to do)
And it is the most frequent case of things failing on their own with modern electronic. And ironically one of the last to think about when diagnosing (I'm also guilty of overlooking that :-) )...

I think here is described what to look for: faint line circles around leads in solder joints are fatigue crack forming. Most prone are heavy components, but small regular ones aren't immune either, mainly when they have short stiff leads not allowing any give for thermal expansion of the board.
Reflowing them using quality flux and solder tends to solve the problem.
Of course be on watch for broken traces. If nothing exploded yet and mainly when the ballast is working at least sometimes, there is high chance the components are still good so the fix could be successful.

I also came across a listing for a Robertson L48 ballast. The seller claims it is new. I'm not sure if it is made in the USA or not, but from what I've seen, the Chinese-made Robertson chokes have a painted finish and say they are made in China on the main label. This one is unfinished and does not say it is Chinese made on the label. Is it actually made in the USA? Also, is it new? It does have some scuff marks on it.

https://ebay.us/m/adDftI

Chokes last a long time so new or used, you'll get long life from whatever you choose. Get whatever is preferable to you. The new Chinese-made one does sound nicer because you don't know how many hours the used one might have on it.