I do think the degree to which software builds on itself is different than in other engineering disciplines, but I'm afraid I don't have evidence to back it up. The closest I can get in the time I've got this morning: consider the amount of (compiled) code Dell ships on a $500 PC and compare that to a $500 mechanical device like a bicycle. I can imagine doing the research to figure out whether the bicycle infringes any patents. With the PC, I can't imagine how it would be possible. You'd have to understand how all of every program shipped on the PC works; I doubt that there is anyone alive who has even read more than half of the code in any single modern operating system. You'd need a team of programmer/lawyers and the source code from every vendor who ships code on the PC.
From there, I'd say that a system where an inventor can't assess whether their invention is infringing without an effort comparable to the amount of work needed to create the invention is a bad system.