Probably something wrong with GENDAC or connector itself. Judging from slightly corrupted text, likely former.

Wow, that looks like the effect I expect to obtain if I flip on the "256 color mode" bit in the Attribute Controller while in text mode. It would turn a monochrome 8x16 character cell into a 4-color 4x16 color cell.

If the problem is that the 256-color mode enable bit is stuck, it would be a defect inside the ET4000 chip. The 256-color mode color chart should look perfect. Before jumping to conclusions, please first make sure nothing interferes on the VL bus. Replug the ET4000 card. Remove any VL IDE controller. Try a lower FSB.

I have two Hercules VLB cards. One exactly like yours, and the other is a Terminator 64 with Trio64 chip. They both developed the same problem with age, cold solder joints on the DRAM chips. It's possible you have a bad joint on the RAMDAC. Try pressing down on it while the system is running to see if the colours return to normal.

Touching chips had no effect.
It happens on at least two different motherboards and without any other cards in the system.
It is not text mode specific. Seems to be affecting all modes. This is what happens when starting Windows (standard VGA mode and default colour scheme):

A hint of blue data is there but with a ghosting effect. Almost like colour covergence misaligment on old CRT monitors (but this is on LCD).

Still unsure if this looks more like a digital or analogue part issue and how an issue like this can develop...

Is video card jumpered? Might had lost the jumper block.

Cheers,

The problem is in the signal path between the DAC and the connector. Check the whole path for the red color. I suspect the choke (the big black round B1-B5 next to the plug)

EDIT
I took a closer look to your pictures. It seems there is more wrong than a missing color. So the root cause seems to be at the digital side. But it cannot be wrong to check the three analog lines from the DAC to the plug. There is a chance that it could be a broken connection and a shifted filter behavior.

It's difficult to find out whether the problem is caused on the analog or the digital side, because the analog signal is sampled by the monitor and displayed in a digital way. If you happen to have a good CRT at hand, the picture on the CRT might help finding out whether the problem is analog or digital. Further hints can be obtained from

• The behaviour of low-resolution modes like 40-column text ("mode co40").
• The behaviour at different color depths (all you have shown so far seems to be 16-color pictures). 256 colors and high-color images would help. In case of a digital fault, the fault might depend on color depth.
• The behaviour of the 1028x1024/256colors mode at 72 Hz (if your LCD can sync it). This mode uses a 16-bit data path between the ET4000W32 and the GENDAC, whereas lower resolution modes might use the legacy 8-bit path.

On the picture you posted it looks ever so slightly like R21 (next to the RAMDAC) was cracked.
I believe R21 is connected to the blue signal however so it is probably just a scratch and not related to your issue but still it may be worth checking it out.