There was a lot of discussion around the time it was being removed[0], but it seems the glibc people simply didn't want the support burden and Emacs was the single user of that interface. I think the issues could have been resolved by checking if all shared libraries are exactly the same as when unexec was called, that the kernel is the same and that the CPU is the same, and failing if anything doesn't match, thus falling back on slow startup. That said, portably serialising the state and loading it back is a better solution IMO.
[0] https://lwn.net/Articles/673724/