Jen,
I guess I should be a little more specific.  There are a lot of variables that determine what a particular environment can handle in terms of file sizes.  These can include number of processors, amount of JVM allocated memory, and how optimized your flow/java services are coded.
The nice thing to know about $xmldata that the system assumes that the file is “xml” and converts the data to a node automatically.  It skips validation/parsing routines as a results which makes your processing in a high-volume transaction environment much faster.  Validation definitely drags down the process.
There is some information relating to this in the large document file handling document.
Ray
#Integration-Server-and-ESB#Flow-and-Java-services#webMethods