I would suggest looking at the waveform on an oscilloscope by inserting bipolar capacitance. Maybe it would be closer to the answer.

... and I assume that you really mean C9 and not C2!

yes indeed!   I had TR2 in my brain... C9 & TR2 - must be the most common EP failure!

As Allf points out: these days you can get large-value bipolar capacitors. I don't think you could in the 1980s when the EP was designed!

I broke my EP getting the 'scope trace above so I will try and get a bipolar and give it a go when I fix it. (Between power-ons, I noticed that C9 held its charge for a long time. So I discharged it before getting a 'scope trace. Unfortunately one time (the last time ) I discharged C9 while the EP was still switched on resulting in the usual TR2 problem!)

If you google "joule thief" you will get a lot more information on how this circuit works. But I could not find any other examples with the TR3/C9/R15/R16 additions - I think they are an "EP special".

So, by using a bipolar capacitor we can improve the situation?

It is possible a bipolar capacitor would make the circuit more reliable... but not yet tried!

I look forward to hearing your results!

I can see that Nichicon makes a 100uf 10V bi-polar electrolytic (6.3mm x 12.5mm), part number UVP1A101MED ... but it is radial and not axial. Is there room for a radial capacitor in the C9 location?
 

I look forward to hearing your results!

I can see that Nichicon makes a 100uf 10V bi-polar electrolytic (6.3mm x 12.5mm), part number UVP1A101MED ... but it is radial and not axial. Is there room for a radial capacitor in the C9 location?

The radial capacitor can also be perfectly soldered. It must be laid.
Sorry for the Google translation...

[ Guests cannot view attachments ]

...the original is 100uF 6.3v but I could not find a 100uF 6.3v bipolar axial...

Sorry, but the original is 100uF 10v ... 6.3v in the schematic drawing, but 100uF 10v in reality.

[ Guests cannot view attachments ]

I ordered these. If you solder it horizontally, I think they will be appropriate ... Size(mm)  L 16 x D 10

[ Guests cannot view attachments ]

I can see that Nichicon makes a 100uf 10V bi-polar electrolytic (6.3mm x 12.5mm), part number UVP1A101MED.

I looked again, and I don't think that the UVP1A101MED capacitor would be a good choice because it has a high-ESR ... whoops!  

I ordered these. If you solder it horizontally, I think they will be appropriate ... Size(mm)  L 16 x D 10

Thanks for posting that picture, the Nichicon MUSE series that you found (the UES part numbers) seems like it should be a great choice, even though they are a bit larger than the original Philips capacitor.

I look forward to hearing if you have any problems with physically fitting the new capacitor in that location.

For anyone whose distributor carries them, alternatives in the same Nichicon MUSE series are ...

UES0J101MPM 100uF  6.3V,  8.9mm x 11.5mm
UES1A101MPM 100uF 10.0V, 10.0mm x 12.5mm
UES1E101MPM 100uF 25.0V, 10.0mm x 16.0mm

Thanks. On my EP it was 6.3V, but of course these are prone to failure and might have been replaced in the past, so it might not have been the original original!

Well, the motherboard schematics do show a 6.3V rating for the 100uF C9 capacitor, and from the voltage readings that you took, a 6.3V capacitor *should* be fine (with my limited understanding of safety margins).

Now the 100uF capacitors at C11 and C12 are there to help smooth out the +5V power supply lines, and if you apply the normal x2 safety-factor for those, then I can understand why 10V capacitors were used for C11 and C12 instead of the 6.3V capacitors on the schematic.

Then from an ease-of-manufacturing standpoint, if they were buying 10V capacitors for C11 and C12, they would just use the same 10V capacitor for C9, even if it didn't need that rating.

I guess that we won't really know for sure until someone is brave enough to try a 6.3V bi-polar for C9!