What did you find? My own fairly casual tests (fine grain film, high quality lens, test charts) found no detectable difference until I tried 3mm window glass. Ctein, a better experimentalist than I, found the same. It sounds as if you have tested fairly extensively and I'd be interested to hear your results.
Cheers,
Roger
Roger were you and Ctein using a microscope to examine the film? I found that with longer focal length or telephoto lenses the only way to go is glass, that the plastic filters cause a noticeable loss of sharpness, no need for a microscope on this one. you can see the difference on an 8x10 print. The longer the lens the worse the degradation. I'm surprised you didn't know about this one, I thought it was fairly common knowledge.
The thicker plastic filters have an edge in plano parallelness but higher thickness also seems to shift focus slightly more, and I would suggest checking focus with the filter in place. The thin filters seemed to have less of a focus shift but required something to hold them more rigidly as they can curve. Both the thick and thin plastic filters have issues regarding being plano parallel to the lens as they are usually in slip in filter holders and are often not parallel to the lens. Using normals or wide angle lenses you would not notice much of a difference. However if you do carry and use teles/longs, then you need to use glass filters, and as long as you're carrying glass filters for your long lenses you might as well use them on everything.
One of the things that made plano parallelness more obvious to me was the use of filters during enlarging. I used to add and subtract various diffusers while printing and noticed not just a diffusion affect but a stretching of the grain at the edges of the print and increased bluriness, more so towards the edges, the more often the filters were swapped on and off during the different contrast (split filtering) exposures. ( I might use a diffuser during the magenta exposure but none during the yellow, etc) I then decided to have my grain focuser lock in on some tiny speck that was obvious. Through the grain focuser a box is visible, I would put that speck on the corner of the box and then add or remove filters, the speck would move all over the place. Now when I do split diffusing, during the non diffusing exposures I add a haze filter to the yellow exposure to substitute for the added glass of the removed diffuser filter, the replaced glass seems to keep refocusing that speck into the same place. I also found that wider angle enlarging lenses had more of the stretching issue than longer enlarging lenses.
I also tested various shutters at various speeds and the intuitive guess that the larger the shutter the more the shutter vibration is correct. Using telephotos, as they tend to exacerbate vibration problems I tend to use them for any kind of vibration testing, I found that at the same speeds, like the 1/4, 1/8, 1/15 range, the smaller the shutter the better the result. The worst offender being the Sinar Auto Aperture shutter which I consider useless for out door work because it's vibration problem was severe. ( a pity though because I own 2 of them and at the time all of my lenses were on DB mounts, I have since converted nearly all my lenses to copals)
I have been curious about whether multi coated lenses, in which the front element has a substantial curve, would have less flare if a high quality multi coated filter were used. My assumption is that multi coating causes what would be considered stray, flare inducing light, that is light hitting the coating from an angle to be reflected. With a flat filter, light hitting the coating dead straight, that is 90 degrees to the coating on goes through with little loss, little coming from progressing angles get increasingly reflected. With a curved lens surface, the curve makes for a huge area in which the 90 degree to the coating follows the curve and creates a far larger angle for stray light to not be reflected by the coating. This is all conjecture on my part and may not show on any kind of tests and may be a non issue. Obviously using a multi coated haze or skylight filter alters the color and makes any test results useless. However I have bought a rodenstock protective filter, allegedly coated with the same rodenstock coating used on their lenses, and without any filtering aspects, for me to test.
What is ironic about all of this testing, is that for a great many of my images, all I'm going to do is print the image through a crappy diffusion filter anyway.........