A lot of it seems like fun, but I am not sure if it is mature yet. My progress was very slow because nothing comes natural to me anymore. At the moment I don't know how "save" cascades in GORM. Maybe I am just not made for the Rails type of programming: seems I want to be in control of everything. With Hibernate, I can explicitly define the cascades, plus it is very easy to find out what to do in the documentation.
To be fair you can probably fall back to configuring Hibernate directly with Grails. I think I would rather have a classic Spring/Hibernate application and replace some parts with Groovy/Grails than the other way round. One thing I also don't like is that Grails doesn't use Maven (or gaven or whatever they will call the groovy replacement for maven they will surely eventually create). I am used to not putting binary libraries into subversion any more and don't want to go back.
My other source of frustration is that I couldn't get my grails application to run in Eclipse and therefore I couldn't debug. Also the Groovy plugin is extremly basic, all it does is syntax highlighting, for all I can tell (and debugging, if it would work). Also, Grails produces the longest exception stacktraces I have ever seen, looking for the source of the error in there is no fun.
I feel a bit pathetic - surely I should try harder and dig deeper into Grails, but somehow I can't motivate myself to do it.
That said, I still think it would be 1000 times more fun than classical Java development. I guess I'll just wait for a few version jumps and then I'll try it again. Wish I could find the time to contribute to grails myself.
Personally I prefer letting a framework do its thing and make intelligent defaults, that can hopefully be overridden (usually the case in Rails).