Hopefully if more people write to the NRMP, the may change their policy, at least for subsequent years. I think it is unfair that someone who could deservedly be completely matched into a lower ranked categorical program should be left hanging dry for a scramble.
Although it sounds like this would be easy and help people, it's actually much more complex than you would think.
One solution is this: have a "rank code" for "Skip to the next categorical match" that you can list on your prelim ROL's. This would allow you to selectively choose what to do when you reach the end of a prelim ROL (after tentatively matching into a rads spot) -- i.e. choosing to skip to your next advanced match, or keep what you've got and scramble for a prelim.
Seems straightforward, right?
Two big problems:
1. The first is recognizing that you could lose everything this way. If you chose to drop an advanced match because you failed to match prelim and try to match at your next spot, there is some chance that you would not match at any more advanced spots (or, alternatively, would not match at any prelim spots). hence, you could end up with only a prelim, or nothing at all, when you could have had a radiology spot. From the NRMP's perspective, just as many people could be upset by that outcome as are upset by the current problems.
You could argue that people have the choice (i.e. to try to skip to their next or keep what they have) but some number of people will "screw it up", and the more complicated it is the more will screw it up.
2. The more concerning problem is the way the match algorithm works. Warning, geeky explanation follows. The match is currently run as two separate matches. First, the advanced/categorical match is run and finalized -- those rank lists are fixed and don't change. Then, the prelim match is run -- for those that matched advanced the appropriate prelim list is used, again the match lists don't change during the match. This makes the match algorithm linear -- it always gets the same answer in a reasonable amount of time no matter what order it is run in. By "order", the algorithm has to take applicants one at a time and try to match them into programs. No matter what order the applicants are processed, you get the same answer -- it doesn't matter whether your match is first or last.
If we try to do what you are suggesting, then the Cat/Adv and Prelim matches need to run simultaneously, and the prelim ROL's will change as the Cat/Adv match is processed. This leads to a non-linear problem -- one that might have several different answers depending on how it is run.
An example may help: Let's say you are trying to match to Rads and Prelim. The way the match is currently run, rads goes first. The match trys to put you in your first rank, but all those spots are full with people who are higher on that programs ROL than you, so you're out of luck. So, the match puts you in your second rank. As the match progresses, there is NO CHANCE that a spot will open up at your first rank. Some people may get bumped out by others higher on the programs rank list, but never by someone lower.
If we try to run the Adv/Cat and Prelim matches together, this falls apart. Same scenario. The match tries to put you in your #1, but all spots are full, so on to your #2. But, then later in the match some applicant who has matched to your #1 gets bumped out of all of their prelims, has chosen this "skip" option, and moves down their list leaving a spot open. How would we decide whom would get this spot? It would require going "backwards" in the match algorithm -- relooking at whether you could match there. If you do, then that opens a spot at your #2 which generates the same problem.
This creates a circular/recursive/non linear problem that is much, much harder to solve. This doesn't mean that it's impossible to solve, but it does mean that there may be several possible answers depending on where you start. These answers would only be a tiny bit different (likely 99% of people would end up exactly the same) but it could create all sorts of problems for the NRMP.
So, this is probably not going to happen, no matter how many people write. Feel free to try, though.