I wonder if the exception was do to the wholesale conversion from the ByteArrayOutputStream buffer to a String. I haven't cracked 10MB yet, so I can't say if mine will have the same problem, but I'm doing that conversion a piece at a time (via the Reader/StringBuilder arrangement). More simply, I'm reading textual characters a line at a time, rather than bytes 1024 at a time, into my output "thing". I'll definitely post back if/when I cross that threshold though.

I wrote a similar function a couple years ago:
http://blog.pengoworks.com/blogger/index.cfm?action=blog:501

You're using a StringBuffer and I used a ByteArrayOutputStream, but I ran into some Buffer Overflow errors if the GZIP'ed content was too large.

In my case I was reading in an XML packet that had been GZIP'ed and the XML files could get pretty large.

Just something you might want to test for…