The exciting thing about CAR-T is that it (academically, at least) enables what needs to happen for cancer to be cured - i.e., a 100% kill rate where treatment (whether that's surgery or chemo) mops up 99.99% of the cancer, and the immune system is responsible for mopping up the last .01% on its own.
Highly anaplastic cancer cells often have ways of being able to evade detection. One of the big ways is via downregulation of MHC-1 (basically a fingerprint receptor that the immune system can use to determine if a cell is acting normally). CAR-T cells can still glom on thanks to their engineered receptor, even if the cancer cells downregulate MHC. This works really well for blood cancers because they are all clonal - a plasma cell in the marrow becomes abnormal and churns out a bunch of cells that all have the same genetic makeup. So if that cancer's main immune escape mechanism is MHC-1 downregulation, it has no effect on the CAR-T cells. Solid tumors are a harder target because the cells often have different mutations in different cells despite coming from one contiguous area.
The two main limitations of CAR are currently:
1) The chance that the CAR cells themselves may become cancerous (although that risk appears to be low...thus far)
2) That they won't persist long enough in the body to complete the 100% killing blow - the current focus is trying to find ways to make them more persistent