That said, I think the various lisps out there have plenty of user—and by that I mean programmer—success stories. My favorite right now is Cascalog. It's an amazing, life changing tool for anyone who needs to analyze data sets using Hadoop. It makes something that would otherwise be too difficult (given the time I have) to even try to do into something that is fun. Nathan Marz, Cascalog's author, should be considered for a MacArthur genius grant.
Cascalog, incidentally, would not be nearly so amazing if not for the relatively few macros that tie everything together.
I don't think this stuff, while brilliant and clever and useful, qualifies for a MacArthur grant when one is standing on the shoulders of N CS giants reimplementing their ideas with few truly novel improvements.
MaCarthy, Kernigan, Thompson, Richie, McIlroy etc are those giants.
I think a lot of people would still balk at the syntax though, and prefer something with similar functionality and more comfortable syntax, even if it was a lot more work to implement such a thing.
Think of the analog of writing a desktop application: there are many object-oriented desktop app frameworks where you can write pretty rich apps exclusively by instantiating existing classes, and this is possible because some domain experts designed a framework for writing desktop apps.
Macros—which I've been using as a shorthand for "syntactic innovation"—are the same way. Someone writes e.g. a web application framework that consists of some macros, perhaps of deep skankiness, that let you write clean, obvious code that solves whatever problem in the web application domain that you're currently confronted with.