I think I will de-lurk here.
I have a reference GTS. I don't think my particular card has as poor 2D that others complain about, although I get some "ghosting" when I use my Linksys KVM switch at high resolutions. I thought perhaps the output filters on my card were designed poorly (as the article suggests) so I took all the components off, measured them, and simulated the filter via HP ADS.
I have all the plots, component values, and my updated filter values in a file here at work, I'll try posting it somewhere.
Summary:
The simulation of measured filter values looked pretty poor, and had a cutoff at around 150MHz (a bit too low based on theoretical bandwidths for 1280x1024x85Hz and above). I then modified the filters to obtain a flat response out to 250MHz (I was hesitant to short the filters as described in the article, who knows what VHF/UHF crap those chips will spit into your monitor).
Results? None. 2D looked the same without the KVM switch (recall that my card doesn't have blurry text.... it's as "unblurry" as a GTS can get). Still see the same ghosting with the KVM probably an impedance issue since the KVM cables are pretty long.
I repeated this with an older TnT card I had lying about. It's original filter was wide enough to begin with. I modified it to flatted the passband a bit, but noticed no improvements in its somewhat blurry text at 1280x1024xwhatever_it_can_do
My conclusions?
Without having a card that was excessively blurry, I can't really draw a conclusion. The RGB output filters would have to be seriously misaligned to blur the text since my GTS filters were already rolling off at 150MHz and I could run my GTS at 1600x1200x75Hz with decent (not a g400, but equal to my V3 on my Viewsonic G790) performance.
I do know that the TnT blurriness wasn't due to the filters.