Whilst this might be true in general melanoma does respond particularly well to immunotherapy with 5 year survival rates in excess of 50% for advanced (metastatic) disease and a subset of patients effectively cured.

I'm just a layperson in this and could be wrong here, but my understanding is that cancer tends to survive by accumulating complementary pathogenic properties that individually are of little threat to the body. One of these properties is increased expression of ‘programmed death ligand 1’ (PD-L1), a surface protein on the cancer cell that binds with ‘programmed death protein 1’, a person on the surface of T-cells that, when bound, inhibit the T-cell response. I could be wrong here as well, but most immunotherapies today are of the 'checkpoint inhibitor' variety, which interfere with this binding process in one way or another.

To me this is similar to removing an invisibility cloak from the cancer cells. Now the immune system gets a shot at these cells b/c there's nothing indicating otherwise. In the case of some cancers, like melanomas and some lung cancers they may look sufficiently broken that the immune system just kills them naturally. But if the cancer cells still resemble healthy tissue, it's not super clear to me what is going to provoke a kill response.

I do think it's likely we will ultimately have customized therapies in which cancer cells are extracted, unique features are identified and custom mRNA packages created to emulate those features sufficiently to provoke the immune system to kill them. That, in combination with checkpoint inhibitors, would likely create an effective response.

I wonder how much of this rate is affected by patients who try immunotherapies have failed frontline treatments and then try an immunotherapy later.