True - scanning of non-infectable files is unecessary with "today's" regime of malware...who knows about future attack vectors/methodologies...
Performing a full scan - on whatever files one deems reasonable to scan versus skip - should only be necessary "one" time provided all future incoming file entry points are monitored and scanned in real-time thereafter. Inarguably, the real-time resource demand is reduced dramatically.
In fact, assuming no new incoming file activity, there would be "no" real-time scanning taking place. Thus, the resource demand would be limited to just monitoring the entry points for triggering activity alone and file/CPU activity would be free to take place essentially as if there was zero interference.
From a hardware perspective, one really need only know that if file activity is originating from the hard drive then no real-time scanning would be needed. So, really the monitoring of the entry points is done via the "reverse" or "inverse" so to speak. A simple on/off switch could be optionally employed to "exclude" real-time scanning of originating activity from the hard drives. In fact, from a developers/administrators point of view, the switch could be automatically changed upon the updating of signature files....or better yet, the full scan scheduler could also be triggered when signature files are updated....hmmm, that could be a feature option regardless of whether this approach is employed or not.
Checksums of previously scanned files still requires in real-time activity to read 2 checksums and then execute a compare operation...admittedly a quick operation versus actually scanning the file - it is still an operation that is not really required. If the hard drive has been scanned then "only" new saves to the hard drive need "oversight"....a copy from one location on a pre-scanned drive to another location on the drive does not need scanning prior to writing. If memory serves me right, I believe that the early implementations of Norton AV utilized something akin to checksums called "inoculation". Certainly analagous if not similar.
Frankly, (or if your name is not Frank then: Seriously

) what is going to be found by the scanner in a file that the same scanner has already found as "clean".....nothing would be my guess

So, the game is just to make sure that nothing "new" ever gets added/saved/stored to the drive/system until it has gone through the scanner...then it is, by definition, safe. Obviously one will want to schedule periodic full scans for the purpose of scanning with the most up to date signature databases - thus, establishing a new clean system benchmark.
It seems to me - IMHO - that paranoria has been driving the anti-malware marketplace to a much greater degree than rational thought. Keep in mind, there is a point of no return with respect to the number of prophylactics employed
...interesting conversation...
BTW: can one join the beta testing for V1.0 ? If so, how?