Performance tuning/improvements

Hi,

I am planning to deploy Velocity(v1.3) in a stand-alone Java application to
process email templates.
A little benchmark test of my own shows that Velocity(with caching enabled)
outperforms the existing homegrown solution.

1) Since the number of emails to be processed are in thousands/hour, was
wondering if there is any scope for performance improvments ?

2) Especially w.r.t reflection, can something be done if the type of the
objects in the context is known and fixed for a given template ?

3) Any precautions that I need to take w.r.t memory usage ?


thanks
Ramesh