Correct me if I'm wrong (since I'm distant from my copy at the moment), but didn't Herd say he had a 'well, duh!' moment about compatibility while demoing something from the +4 family? Of course, once the obvious hit, Engineering then did begin to bug management about it (leading to the 128, which got some weird looks for its chip count).
The impression I got was that the compatibility concept was late to bloom at Commodore not, as we might expect, from mismanagement, but rather from the outright lack of involvement from management or sales that left Engineering to be a computer scientists' playground -- lots of fun, try not to lose an eye. ;-) Of course, since management or sales could rarely articulate a need that made sense, they ended up taking whatever Engineering had to demo and then trying to cram it into a product-shaped hole.
[Or, in short, imagine if Linux's only route to the world was through Lindows.]
