It may be possible to run a zeebe test container per class or per the entire test suite rather than per test, isolating data based on its definition key rather than process ID. As a plus side, this would give us a lot of speed benefits vs. the existing per test container. On the flip side, we have to manage data isolation in tests ourselves, making sure we consider only data from the process that we are testing

Hints:

• We would probably have to remove/refactor the tests that reduce batch size down to 1 during testing, as the data from other tests would potentially still exist in the exported indices
• We could even keep the existing logic with prefixing for the whole test class
• One risk is starting a new test while data from the previous one hasn't been flushed from Zeebe yet, leaving an exported record. Using unique process keys might avoid this being problematic though