I am happy to report, that flares are now working. The bugs you described (full opacity when looking at them over an edge) have been fixed. Fade-in / Fade-out ist working, too.

AWESOME

you should commit these changes to ioq3 itself too as well as release a "flares" patch

I was doing some more "research" for a patch to ioquake3 and noticed an annoying bug that won't appear in EliteForce but in quake3 maps.
The flares can suddenly change their brightness when they are over a surface that is the last one before emptyness, when certain weapons are held / from a certain view angle. This means maps with a full skybox won't have that bug - which is the case in EliteForce, as far as I can tell. But in quake3 there are many fully enclosed maps where the bug will appear.
I will post the patch to the icculus.org mailing list and ask for suggestions, as this seems to be an OpenGL problem.
Furthermore, flares seem to be very expensive on the framerate. It's not really an issue nowadays but I find it curious that so much FPS is lost on these flares.

wierd, perhaps the method taken is just slow? EF's flares were very fast on my k6-2 500 w/ gf2.

Nope, they eat up 20 fps on my bro's original, too.
Have a look at ctf_neptune.

EDIT: I gotcha some comparison screenshots:
http://thilo.kickchat.com/efport-progress/screenshots/noflares.png
vs.
http://thilo.kickchat.com/efport-progress/screenshots/flares.png

The algorithm to get the size of the flares is a bit different. With increasing distance, the flare size stays nearly the same proportionally to the screen size. It is the intensity that will fall off proportionally to 1/distance^2
That's the way the guys at ID originally wanted it to have. In EF it is simply a kind of sprite.

If speed's the problem then maybe have an option for the EF/q3t 1.05 method, then? I don't remember intensity fading by distance though, not sure.  maybe have a glReadPixels based flare rather than traceline?

also ef's flares fade out behind 1-bit alpha masks too, which is easily seen in The Machine HM level, and that probably uses glReadPixels too.

(the only card flares "broke" on i've ever used in the past was an s3 savage2000)

Nope, I don't think it will be better using any version.

To the bug I mean:
The two pictures show the same flare from exactly the same position/angle. The only thing different is the weapon.
The weird thing is, the flare is considerably darker when holding the lightning gun than the rocket launcher.

This is normal intensity:
http://thilo.kickchat.com/bilder/screenshots/flarebug-normal.png
This is buggy intensity:
http://thilo.kickchat.com/bilder/screenshots/flarebug-toodark.png

I have no idea why this f*cking error happens, I'm already hunting it all day. I don't know enough GL, though, so all I was doing up to now was pretty much trial&error.
I checked with my debugger and it seems like the bug only appears when RB_AddFlare is called before all other surfaces have been written. Some modelMatrix and orientation stuff seem to have an impact on how this bug appears, too.
I'm going to bed now, I'll send a patch to the ioq3 mailing list tomorrow and ask the OpenGL gurus there what can be done about this. Because apart from this annoying bug that only affects a few ones on some maps, flares are working nicely.

Very strange bug. Maybe the size of the md3 is causing it, or the traceline check? (Wild guess) Did you try cg_drawGun 0?

tracelines are really slow btw, and i'm really confident that elite force does use glReadPixels (which is faster, but "breaks" on certain hardware). i did a video because screenshots suck lol:
http://www.planetgargoyle.com/vid/efflares2.avi

i'm probably headachy myself and i think im repeating myself somewhat. i need sleep D:

To be honest, I have no idea what these gl calls are doing and what you're talking about.

On the other hand, I gotcha a little video I just captured with ioq3's AVI feature. I reencoded it using mencoder... pretty cool I have to say

http://thilo.kickchat.com/videos/hm_dm1-flares.avi

There's also a patch for enabling flares:
http://thilo.kickchat.com/efport-progress/patches/q3-flarefix.diff

EDIT: I have prepared a couple of videos still:
http://thilo.kickchat.com/videos/q3dm19-flares.avi

This video shows the bug I mentioned:
http://thilo.kickchat.com/videos/q3tourney2_flarebug.avi

btw, Wudan got glReadPixels (elite force style) flares working in q3a:

http://www.quakesrc.org/forums/viewtopic.php?p=59037#59037