I think all the talk of being faster than gmail search is going to last until they actually get a lot of users. It's easy to make search fast for small datasets.
You're not making any sense. They can shard the hell out of their indexes since all searches are user specific, so that's a non-issue. They can afford to throw enough hardware at the problem if their business model works out and they attract enough paying users. It's presumably not cost effective for Google to offer a super snappy search experience through gigabytes of user-specific data, which is why Gmail search sucks so badly for people like me. (Does anyone know if search is way better if you're a paying customer of Google Apps?)
The existence of a way to shard searches doesn't make scaling real time search on email (hint: do some back of the envelope calculation on how much data that involves) a non-issue.
Fair enough, I too readily dismissed your drive-by objection, however non-specific and snide. Scaling real-time anything is never easy. My main point was that cost effectiveness is what keeps GMail from offering a similar level of search quality. If Greplin can get a large enough fraction of users to pay for their service, the technical challenges are surmountable.