Those are great questions. The answer is going to vary (quite dramatically) by the target disease, target cell type, intended effect, etc. For example they mention that 21 percent elimination can 'reduce symptoms', so you could be looking at a situation where 80% elimination slows down the progression enough that the low-symptom phase of the disease lasts long enough that you can die of some other age related complicated instead. Die with the disease instead of die because of the disease - you see that as a common goal in oncology now, and in treating diseases like HIV. Rate of reversion might depend on the cell type.

Add in the aspect that it looks like they want to work with a new AAV vector optimized for the brain (looks like the compactness of this epigenetic editor is one of the advantages which tracks, can fit in an AAV then). So if there are muliple applications required there's also questions of, what if in future applications there's an immune response against the AAV used that reduces uptake, etc etc.

These questions (along with all the others) tend to be super complicated to answer, which is why all this stuff (more mouse trials, clinical trials, manufacturing, etc) can be so long/complicated/expensive. Clinical trials are also lit in that you boil all the unknown knowns and unknowns into 'let's just check and see if it's safe and if it works.'