I don't know how significant this [1] is a barrier to such treatments, but it seems likely to be more of a potential problem for cures vs. sensors w/potential false positives.
In theory, yes, but delivery of the CRISPR enzyme to all the infected nuclei is still an unsolved problem. Also, it doesn't cut its target every time it does get in - some of the protein is inactive (probably just broken/malformed/misfolded) - but increasing the dose to compensate leads to more instances where it cuts the wrong target.