I'm going to test a transactional queue system on Unix.

Hate to reinvent the wheel, anybody got test cases or ideas on this topic from previous test battles?

I guess it's important to stress this system by randomly seeding it with both small and large queue contents? Any more?