Yes, I could write unit tests (I will think about how to do it easily) for any rule developed in order to test that rules match, however I don't think performance should be assessed through unit testing.
IMHO a realistic performance test should be ran on a running system with real events generated. If you measure performance through unit tests, you will have a performance test in the worst case scenario because only matching events will be tested (a matching event takes more time to process than a non-matching one). If you want to measure the maximum throughput of events that Gene (and WHIDS by extension) can process, you can use the -progress command line switch in gene command line. According to my tests, for a bunch of 100 rules ran on 60000 events collected on a running machine, the engine is able to process around 7000 events per second when ran on a single job (-job switch). This way we can assess the maximum throughput the tool can handle in a realistic scenario.