I flipped through and there doesn't seam to be anything about weak hash tables, but that's not surprising seeing as TAOCP is written around an assembly language, while weak hash tables make most sense in a GCed environment.

If they referenced it and the patent was still granted, that's bad. It means that the USPTO still considered the patent to be "novel" in spite of all the information found in TAOCP.

I put "novel" in quotes because the courts seem to use a hair-splitting definition of it. I honestly think that, when faced with a pile of Lego blocks, there are patent lawyers who could argue that it's non-obvious that one could combine them.