This came out of Gentry's PhD thesis under Dan Boneh, for which Gentry won the ACM Doctoral Dissertation and Grace Murray Hopper awards, and later the MacArthur Prize. It was the first demonstration of fully homomorphic encryption, and therefore resolved a very large open problem in theoretical cryptography.
Depends on the scheme and your security parameters. The best known ones we have are probably 5-6 orders of magnitude slower than plaintext in the best case (obviously the worst case is much worse).
Also, how much slower both in terms of throughput and latency. Low latency Low throughput is not so bad for some applications as would high latency high throughput for others.
More than a few. Also, all the benchmark algorithms I've seen in papers have been trivial so this is very much still an active area of research. The idea is tantalizing though, and I'm looking forward to new breakthroughs in this space.