Well guess what? With 75ohm resistors on each R, G and B line tied to ground it syncs in 640x480, 15khz every time connected to the arcade monitor! EVERY time!!
That's very good news! At least we now know it's a problem related to monitor detection. I think I already asked but do the other output have the same problem?
I guess it does prove it absolutely needs EDID info if this is what the resistors are supposed to circumvent.
Not at all. It does prove it is related to monitor detection. EDID info is a separate thing and not required at all as you're seeing. I do know this because I have a Soft-15kHz EDID dongle and it doesn't help when it comes to monitor detection.
The other problem I am having of course is the issue with MAME and resolutions/refresh rates. Right now I have a ton of resolutions available but all are 60p so none of the refresh rates in MAME are correct. Your suggestion of the "super" resolutions only gave me a handfull of interlaced modes, no progressive modes or accurate refresh rates.
Aren't you using GroovyMAME? Those modes are supposed to be used with it. GroovyMAME will recalculate the required refresh on the fly, you'll have the right refresh for all games. On the other hand the "super" resolutions just work, do check the modelines list to see what it's doing.
Now, in case you wanted to test any further: before you put the resistors, the issue seems to be (as it was in Sledge's case) that the system can't see the interlaced modes on boot (because your monitor is not detected and the system thinks they're not safe), so setting for instance 640x480@60i will be automatically reverted to 1024x768*. But I'd say that if you tried to set a 15 kHz progressive mode like 640x240 (or whatever other that you have but not a 31 kHz one!) with the new option "set as default display mode" maybe it would stick (in case it's working as it is supposed).
*1024x768 is the default resolution in W7. However you can force W7 to use a lower resolution by default. From an elevated cmd, run "bcdedit /set vga on" and reboot. At least this might force the system to fall back to a mode that is readable throught the j-pac.
BTW the EDID overrides wouldn't help here either (I might be wrong, if so let me know). The problem happens deep inside the display driver when it doesn't detect that a monitor is connected. The EDID override wouldn't be considered in this situation.