I haven't seen evidence of cancers re-mutating mutations. They tend to develop new mutations (as in the clonal evolution hypothesis), and they can completely change their cell surface expression. But I suspect that they are unlikely to mutate an already mutated gene.

I guess for the treatment to be most effective you need to target some of the earlier mutations, rather than a small clone.

From my perspective, and not having read the paper, I thought the technology to predict what TCR would bind a particular peptide on a particular MHC-1 was not there yet.