The problem is, of those 600 packages:
25% are Moose extensions, or for testing Moose itself.
25% are Moose components of large frameworks like Catalyst, DBI & Mason.
Another 25% of them have specialist dependencies like Amazon S3, Hadoop, Twitter, Jabber, KiokuDB, POE.
Of those remaining, they are mostly peoples own experiments that are neither examples of good Moose practice, nor incomplete. Eg. Acme::Mahjong.
It seems that despite three plus years and all the hype, it's mostly Moose (and one or two in-house projects) that is using Moose.
And that can possibly be attributed to the complete dearth of examples of applications that use Moose to do real work, rather than cutesy examples to demonstrate it feature set.