I've used Diafine for Tri-X in 35 mm a fair amount, and also with diluted Bath A for Copex Rapid microfilm in 16 mm. I've never seen "posterization" in my scans with Tri-X, though I have seen it with microfilm that was underexposed. The image you point to is one Lex posted as an example of Diafine's effect at the grain clump level -- comparing Tri-X at 1250 in Diafine against (IIRC) Delta 3200 shot at 1600 in DDX (Tri-X in Diafine wins, IMO). He was using an older 35 mm scanner that didn't really like B&W, also, so it's very possible the effect you're seeing is due to the scanner.

Here are a couple examples of my own shots done in Diafine (Tri-X, 35 mm, shot at EI 1600):

http://www.photo.net/photodb/photo?photo_id=2442154
http://www.photo.net/photodb/photo?photo_id=2442163
http://www.photo.net/photodb/photo?photo_id=2513728

I don't see any indication of posterization in these -- and I'd be happy to send you a 1:1 crop from the full 2400 ppi scan of any one for examination. If you look in my portfolio over there, you'll see a number of other shots done on Tri-X in Diafine -- I challenge you to pick them, via tonality, from TMY in HC-110 or Tri-X in Caffenol.

What might have happened with Lex's image is that, shot in bright daylight, all the image information would have been compressed into the contracted shoulder, and a histogram stretch after scanning to get good contrast could lead to an artifact from using only, say, 4 bits of the 8-bit gray scale available. This doesn't happen in printing; nor does it happen with a scanner that has a longer native sensitivity range than the antique Lex was using when that image was posted. My Arcus 1200 uses 12 bits per channel internally and has only exhibited this effect with microfilm negatives showing a very, very narrow dynamic range.