I like to collect fluorescent lamps, even used ones that ordinary people discarded, for at least 2 reasons: First, they can be used for high voltage experiments as long as the gas filling is intact (mostly only the electrodes are worn) and second, discarding  them as regular waste is prohibited so I'm actually doing something good by taking them home from trash and preventing them breaking. I have also found 3 of them near utility poles.
With high voltage most of them work just as good and brand new lamps.
I have an interesting specimen in which the mercury was absorbed and only the gas filling is excited. Originally it was a 4000 K CFL, now it light up red.
/>www.youtube.com/watch?v=WrVs6QCzuW8
www.youtube.com/watch?v=pr0yquq6FK8
https://www.youtube.com/shorts/Uf-0d5U2ii4

Some side notes if you want to try this: The electrodes of worn lamps get hot and sometimes can crack the glass. It it better to power them by capacitive or inductive coupling.
Do not let a spark jump to the glass tube, it can puncture it.

here are some photos of my setup

Photos always help.

Next step is to scope signals at start of enable pulse.   Scope both GDT inputs (UD1.3 outputs).  Scope CT input.  That will likely be enough to show issue.

Wrong GDT core material would be another possible cause for your symptoms.

Yeah probably overheating, hopefully the new bridge will be able to handle it. It was drawing ~20kW and sounded really good.

The toroid heated up so much that it collapsed by a few millimeters (toroid is 3d printed).

Is there space between top of secondary winding and toroid?  If little or no space, induction heating of toroid is likely cause of heating (and of wasted power).  Magnetic field lines need a path to return to bottom.  (I've built a couple coils with ferrite plates between secondary top and toroid to minimize gap needed.  Much easier and lighter to leave a gap.  I'd recommend a gap of roughly secondary radius, half radius at very minimum.

Other less-likely possibility is poor electrical connection between foil tape and secondary top.  Most foil tape has non-conductive (or minimally conductive) adhesive.  Usually there are enough random points of contact at edges.  Sometimes requires folding under a narrow hem on tape edges to contact adjacent foil.

Finally blew up an igbt running 1500A OCD, 600Vbus, 25% dtc, 1ms on times. I think it's time for a 1200A bridge.