I think this “third parent” thing is a bit overstated. 49.99% from mom and 49.99% from mom and some 16,000 bp of mitochondrial sequence out of some 3 billion from a donor.
Eventually I’m sure it will be possible to replace defective alleles on nuclear DNA — in most cases probably with a gene from whichever parent has a healthy version (unless of course it’s something on the little Y chromosome). But that’s wildly more complicated that replacing mitochondrial DNA. With mitochondrial DNA, they’re replacing the entire thing, which is a whole lot easier than replacing a little piece of very long strand of DNA, especially since it’s entirely outside the nucleus, so that the nuclear DNA doesn’t have to be disturbed at all. They just take the nucleus from the just-fertilized one cell zygote, and transfer it into an enucleated donor egg — this can be done by microsurgical techniques on the zygote and donor egg.
To replace a defective allele on a strand of nuclear DNA, you’d have to go into the nucleus, isolate and uncoil the chromosome where the defect was located, take out the tiny portion where the defect is located, insert a tiny portion containing a normal allele, and do this all in a way that doesn’t damage the strand of the DNA and prevent it from continuing to replicate. Realistically, this will never be doable by microsurgical techniques — it will require extremely customized genetic probes which are inserted into the nucleus, and which automatically seek out and latch onto the specific defective sequence, automatically remove it, automatically replace it with a normal sequence, and automatically reconnect the breaks in the strand which had to made to do the replacement, and then automatically detach from the strand.
This is theoretically doable, but to do it reliably enough to use in human reproduction is likely many years away. The find-and-latch-on step is already in use, but only on sample cells pulled from an embryo for genetic testing, not on a cell which will continue as part of the developing embryo. In other words, we already have the ability to create a custom probe which can be placed into a nucleus, where it will seek out a specific abnormal allele, e.g. Huntington’s Disease, latch onto it and put out a fluorescent glow to show that it has found and latched onto the target defect, thus revealing its presence. But that’s the end of the line for that particular cell. To actually fix the problem would require all the other steps. There may be a few specific defects that could be fixed without total replacement, if the defect involves only a transposition of a couple of nucleotides — I can see a next-generation probe being able to find, latch on, and swap a pair of neighboring nucleotides before disintegrating harmlessly and leaving the repaired allele behind. But for defects where the “misspelling” is more complex, it’s going to take a much more complex process.
Probably well before that technology is developed, we’ll have the technology to identify individual gametes that have a particular genetic defect, and just not use those gametes for fertilization (to a certain extent, there’s a method that can already accomplish this for eggs, but not for sperm). And then we’ll learn to swap out entire chromosomes (which probably can be done microsurgically), so that the defective allele would be replaced when the whole chromosome is replaced.