Multidrug therapy will become the mainstay. It works for HIV (HAART) because despite the high mutation rate of HIV, it is exceedingly rare for the virus to sustain the ~3 different mutations required to confer resistance to 3 different drugs simultaneously.
I imagine that this will, in fact, be how we end up applying our targeted cancer therapies as well.
Let me put it this way: each base of the genome of a {cancer, bacterium, virus} has some chance of mutation at each replication step. If you target just one protein, then you may need as little as one mutation to confer resistance. In a normal human genome, you would expect this to happen to the nucleotide of interest once every hundred thousand cell divisions. Obviously a cancer has a higher error rate and millions of rapidly dividing cells, so this would occur quickly. If you target two proteins, you would expect both mutations to co-occur once every (100,000 x 100,000) cell divisions. If you target three proteins... Now you are getting somewhere.
Certainly it is not quite this simple. If variants with one mutation are more fit than those without it, then resistance to one drug in the cocktail will be selected for. But with HAART, we have seen that the multidrug approach is fairly - perhaps surprisingly - robust to that possibility.