How do you envision that scaling for say an factor of five increased load?
If load levels continue to grow, leverage ehcache to handle caching directly in the VM (if you're not already), begin splitting responsibilities across different service implementations, etc.