Well, the Exp had more than a single problem actually...
First of all the base was 12.186 polys heavy... can you believe that, but the model is way lighter.
The cvh tho... 10.812 polys... GAH wtf i thought
look for yourself
somehow orion's exporting script went awry or somethng...
this is how it should look like:
It still didn't work, so i have redrawn the cvh almost entirely. In the end the second problem was again a radius problem. the cvh reported a radius of 3.6+ while mdlview calculated a 4.9
That said, this CVH has been tested and it works, it's quite precise, doors work, no more entering sides or back or anything, and you bump where you should.
http://dl.getdropbox.com/u/1790605/rarfiles/ss400.cvh
I used the original artwork model so...
I include these...
http://dl.getdropbox.com/u/1790605/rarfiles/gtpali.rar
http://dl.getdropbox.com/u/1790605/rarfiles/gtres.rar
http://dl.getdropbox.com/u/1790605/rarfiles/rixexp.rar
http://dl.getdropbox.com/u/1790605/rarfiles/gtgarr.rar
the ones i worked on earlier. GT pali and res were tested by me and kramari and worked fine enough, the rix exp has ben tested by me only but it works (it's worth a look at th cvh at least i hope, but i think compeller may have a better work already, cause mine was done in a hurry)
The Gt garr isn't perfect imho, but i tried it out til it seemed to work well enough (no bumps while flying by the pylon corners, no bumps halfway through the door and such things)
all of these were done before we found out about the scaling range issue. That means they're customized to work well enough but i've done it by tailoring the cvh in milk instead than checking the radius in the cvh proper. Noy a veryelegant way to solve those issues i know, but i urge you to at least try the files before trashing the whole thing, it may be that they work well enough to be used sparing the people working on doors and cvh a bit of time.
P.S. the GT exp fix was done by checking the radius, so it was made "properly" oh and it's 1797 faces against the original 12k+ not bad huh)
Edit: strike this too, if we don't count in a weird cvh poly number the new cvh accounts for more or less the same number of faces, so you won't really notice any big difference, except that it's maybe a little more snug-fitting
Hope this helps.