What would happen if the proofing system occasionally *inserted* an error into the page and the double-checked that the known error had been found and fixed? eg: find a correctly spelled word with "m" in it and change to "rn". Choose from amongst a list of 100 similar things.
The problems I see is that it would be hard for the system to "model" the kinds of errors that remain unseen in the book, thus it would train proofers to look for the wrong things. Also, if the system can introduce errors, it had better know how to take them back out. And its not fair to introduce errors on pages that a particular individual has already proofed, for example if for a given page I P2 and PP then I do NOT want to have to go into "paranoid mode" and look during PP for errors introduced after I had already P2'ed. Proofing *already* leads too much to the feeling that one is chasing one's tail, going around in circles, and "didn't I already fix that one already!"