Unless you can conduct this experiment over millions of years, it won't work. In fact, it won't work even if you do conduct it over millions of years. But that's evolution's dirty little secret! :-)
Let's say the average fruit fly lives a week...from egg to fly. In one year, in one experiment, we can go through 52 generations of fruit flies. Let's say we run the experiment 20 years. That's 20 x 52 or 1040 generations of fruit flies. I don't think evolutionists base their beliefs on time, but rather on generations. Humans have supposedly made great evolutionary leaps over say, the last million years. If a generation of humans is 50 years, then that's only 20,000 generations.
Now surely in 1040 generations of fruitflies it seems highly likely that we should see great evolutionary leaps in our fruit flies who are being force to get food in progressively deeper water. We can even conduct multiple experiments, thousands and thousands, at the same time.
Now surely, in an environment specifically optimized and tuned to favor evolution we should see evolution. After all, it supposedly happens completely by chance in the real world.