How I fixed the backlight of a LCD monitor

WARNING: THERE ARE EXTREMELY HIGH VOLTAGES RUNNING ON THE INVERTER CIRCUITRY, SO BY NO MEANS TRY THIS AT HOME WITHOUT PROPER TRAINING AND CERTIFICATION. I'M NOT RESPONSIBLE IF YOU READ THIS AND THEN BURN DOWN YOUR HOUSE, KILL ALL THE FLEAS OF YOUR DOG OR WORSE, HURT YOURSELF OR ANYBODY DIRECTLY OR INDIRECTLY CAUSED BY YOUR STUPIDITY.

Model: IBM 9516 TFT LCD Color Monitor (got it from the trash, so cost was $0)
Symptoms:
Backlight turning on and off cyclically then off permanently, with a period of roughly 0.5 second.
I could actually see the setup menu image (see figure 1.) during the brief moment the light was on.

Figure 1. OSD Setup Menu for LCD

Then I used a rough cable to see if video circuitry was working, which thankfully was in perfect order,as I could see the computer image in there. While in the midst of these tests, the backlight went off permanently, which kind of put me off.
I opened the monitor back (see figure 2.)

Figure 2: LCD IBM9516 back view

and checked the backlight inverter PCB: found a SMD 1.6A fuse (F2 on figure 3.) which was open.

Figure 3. Inverter Fuse rated at 1.6A

I followed the circuit lines and they went to the high voltage part of the PCB, so suspected something wrong over there. Shorted the fuse with a multimeter (Amp scale) for a few moments while power was ON and the backlight comes back on, but still cycling on and off.
I take out the protective shielding from the high-voltage inverter power driver and one of the SMD power transistors shows a bit of solder out of place. (see figure 4.)

Figure 4. PNP Inverter power BJT

I do a rough check with multimeter in the diode position, reaching the conclusion that the B1203 (PNP 30V 3A) transistor Collector and Emitter terminals are shorted.
Desolder the transistor from the PCB and notice that since the transistor was shifted from its correct placement, the back heatsink (Collector) shorts a PCB line that connects to the Emitter! The SMD transistor probably overheated and melted the solder that kept it on the PCB and it shifted away. The D1803 NPN transistor next to it also was a bit shifted...
Since the monitor has 2 pair of fluorescent lamps, each pair connecting to a power inverter driver , the circuit is mirrored at the left and right of the PCB. (see figure 5.)

Figure 5. Backlight Inverter PCB top face

I turn it on again and predictably none of backlight fluorescent bulbs turns on, but after turning it off, I check the other circuit transistors and one of them is quite warm. I turn it on again and check its terminal voltages, giving 20V at the Emitter and 19V at the Base: strange! Take out the transistor B1203 (PNP) and test it, and it turns out that it's broken.
I take out the remaining tape and shielding from one of the power drivers that is covering the bottom part of the PCB board and interestingly a SMD power resistor (391) terminal comes out, glued to the tape covering it.Fishy! (see figure 6.)

Figure 6. Closeup over the 390ohm bias resistor

I checked the other power driver and the same exact resistor terminal comes out too! Probably a bad soldering mask: I resolder the resistors. The resistor is for sure part of the power transistor biasing circuit:didn't bother to follow the circuit lines to it find out though.
Go hunting for a substitute of the broken B1203: Mouser gives an equivalent as blabla. Get datasheet for parameters and go hunting in the junkbox for high power PNP transistor and get one from a non-working computer motherboard (D45H2A). I solder it to the PCB, first in a crude way, just to check if the circuit works correctly.
I also solder back the transistor that I removed before from the circuit , after testing it out with the multimeter, and it seemed to be in good condition. I was careful enough to make sure that this time the transistor terminal connections weren't being shorted.
Shorting out the broken fuse again the light came back on permanently, so I put everything back together. First I had to proper solder the replacing transistor and so I adjusted the terminals and used duct tape to isolate the collector from a nearby inductor.
 (see figure 7.)

I temporarily "fixed" the open fuse with thin copper wire and used duct tape to isolate the high voltage circuitry shielding. Then I built a VGA cable to connect to the monitor. This monitor requires 5 BNC connectors, 3 for the RGB signals and 2 more for the Vertical and Horizontal Sync Signals.
I then configured X server to allow Xinerama multi-head (actually Nvidia TwinView).
With the fixed IBM monitor, right now I have two LCD monitors connected but only one DVI port. The Nvidia automatic detection mechanism insisted that the DVI one would always be assigned as second monitor, even if I use the"UseDisplayDevice" or change the order of "ConnectedMonitors". Crawling the Nvidia forums , I found that Nvidia recently added a configuration parameter which allows me to select which monitor will be the TwinView/Xinerama primary (first) by using the "TwinViewXineramaInfoOrder" option.
You can take a look at my two LCD monitor X configuration, but be warned that you will need to use (on Linux) at least the 1.0-9x driver version if you want to use the "TwinViewXineramaInfoOrder" option.
Everything set up, I only had to enjoy the nice multi-head!
Tests were made with a crude power supply, using two LM317 linear regulators to give 20V from a 2x12.6V 2A transformer.
ToDo: SMPS with buck-converter 2x12.6=25.2V 2A input (50.4W) 20V 2.3A output (46W)