Everyone's cancer is its own disease (if not several), so the researchers are fighting one disease (or several), instead of an infinite family of diseases. They are not trying to cure pancreatic cancer - they are trying to cure Steveoma. You're absolutely right that this is the mentality that future treatments need to take. In terms of treatment response, a liver tumour could have far more in common with a brain tumour than another liver tumour - the anatomical pigeonholing of cancers is a throwback to pre-molecular diagnostic times (so about 5-10 years ago and before) and is something that will hopefully change over time.
They can find an exploit in Steve's cancer genome on Wednesday, design a molecule to hack it on Thursday, synthesize it on Friday and start titrating it into the patient on Saturday. If this time frame were possible this would be ideal, however, unfortunately each of these steps (if they are possible) takes significantly longer than an evening, and typically far longer than a patient has.
Moreover, Steve isn't on just one drug. He's got thirty or forty teams attacking every vulnerability, theoretical or practical, that may exist in his cancer cells. This is a risky game. How do I (the scientist) know which drugs were effective? How do I know how they'll react with one another? If my focus is Steve, then do I really want to risk killing him through an unexpected cross reaction of my 30-40 untested drugs?
Not everyone is a billionaire. But if you do this for enough billionaires, the common elements in the problem will start repeating and the researchers will learn a repertoire of common hacks. How many billionairs? There are (apparently) 1210 billionairs in the world. So say they all get cancer and get this team of doctors - if any of the patients survive then some aspect of the treatment worked. If they don't some aspect failed. It would be impossible to determine which is which. Considering the heterogeneity in terms of cancer genetics, 1210 would tell you very little at the genetic level.
You'll note that a lot of cancer patients die anyway. But a lot fewer than used to.
The entire process we call "drug development" is an attempt to gain six-sigma confidence that we are not practicing quack medicine. Especially for cancer, do we need all these sigmas? And are we obtaining them in an efficient way? I can't imagine how anyone would even begin to argue the point. This is not the problem with drug discovery (or correct).
[TL/DR] There are aspects of this which ring true (reproducibility is an issue, although how you measure it is also a source of contention, and mouse models have their downside, certainly), however, there are also gross assumptions and misunderstandings which would make this course of action as ineffective as it would be expensive. There is an absolutely imperative need to begin to asses and describe cancers as more related to a patient than a disease title, you are quite right. But to forgo basic research in favour of arbitrary treatment-come-experiments would be to dismiss the last 100 years of biological research which have provide us in a position where you can make your argument. A combined focus of translational and basic research is essential, with a frequent and effective dialogue between the two.
Ultimately, your argument assumes the rate of discovery will be constant. You need to consider the research in the broader picture. Cancer is a genetic disease (in terms of pathological origin, not necessarily in hereditary terms). It cost the (publicly funded) human genome project $3bn to sequence a human genome in about 12 years. Now it costs $5000 in a week or so, and both the cost and the time is coming down. As I said before, it is an exciting time to be a cancer biologist.
-- I can't reply again, but, um, are you trolling?