One thing puzzles me. The social utility of patents is supposed to be that they spread knowledge, because otherwise people would keep their discoveries secret. Well, how many people make use of the "knowledge" in software patents to further their work? When was the last time any of us said, "Gee, this is hard, I'll do a patent search to see if anyone has solved it"? The very question sounds like a joke. In other words, regardless of what one thinks about software patents in theory, in practice their positive uses are negligible compared to their negative ones.
Also, just because you don't read patents doesn't mean nobody does. I would guess it's more of a trickle down effect. University professors and researchers read patents and use it to expand and continue research and gradually for the good patents by the time the patent has expired the knowledge it imparted has diffused into the working participants in the field.
I think it would be great to see the courts adopt that standard. It's there in the law, and repairs the problem of the obviousness standard, which is to say that the patent troll just pays some expert to say "it isn't obvious to me!" and done. But teachability is a much more defense-friendly requirement: the jury typically won't be able to understand it, so if the defense brings a line of smart people who say "this is gibberish" that ought to be very compelling.
Perhaps no one reads the actual patent document (as you say it's written in legal terms - so it's purpose is to establish the extent of the invention for the law). Perhaps what people do do though is look at the outcomes of patents like the devices or code and learn from them. My point is that just because some people (or even no people) read the actual patent document that doesn't mean that the knowledge doesn't reach those who might use it.
The trade of still isn't worth it in our industry at least but the secondary benefits should still be considered in weighing the balance (there are some real benefits of patents along with the considerable costs).