Hey,

So, I think the test suite is pretty widely regarded as being in total shambles. I personally haven't managed to make test_integration work at all, though I have proof-by-existence that the application does in fact work. Additionally, the test suite fails more often than it passes due to dict ordering issues and the snappy tests fail if python-snappy is installed.

Does anyone have any helpful hints about how to make test_integration actually work, or should we just comment it out for now? It seems like most of the problem is in spinning up zookeeper/kafka as external processes, so it might be reasonable to just rely on an existing cluster?

My gut tells me that the right course of action is to disable the integration testing until we have a better solution in place, turn on test coverage reporting, and then start moving tests out of the monolithic test_unit and into their own files. From there we can start looking to improve test coverage. I'll have a pull request going down this path Soon(tm) unless someone objects.