The compatibility test by default tests all possible client/server combinations. This includes testing already released versions against each other, like 10.6.2.1 client against 10.1.2.1 server. Testing for incompatibilities between already released versions isn't of much interest for the ongoing development. For example, when testing a change on trunk, you're really only interested in the combinations that have trunk either on the server side or on the client side. If we could filter out the combinations that didn't test any code on trunk (or some other branch we want to test), the time to run the tests could be significantly reduced.