How are our existing test data inputs and outputs captured?

The under-the-covers techniques used to intercept data flowing through jobs is different for Parallel and Server jobs. In each case, you simply run the job with its 'Unit Testing' parameter to 'Intercept'. This will cause the job to behave normally with the exception that MettleCI intercepts (in memory) data flowing down the input and output links and saves it to the relevant test data files. This data can then be checked in to git and distributed downstream with your job where it can be used to perform tests without requiring access to the original data sources. Let me know if you'd like more details and we can arrange a call.

Please see our documentation of Unit Test Data Interception.