The genes essential for cancer cells to grow in vitro could easily be quite different from those required to establish the tumor micro-environment, invade new tissues and fight off the immune system. They'll have many false negatives from the mismatch in context. Simultaneously, they might have false positives from e.g. differences in layout and density of tumor cells (2D/uniform vs. 3D heterogeneous tumor).

And even if they get the right targets, finding drugs with an effective therapeutic index is hard.

This might be a useful screening technique...but it's not the first high throughput screening method. They might have some successes, only trials will tell.