Excellent thoughts, thank you.

To clarify ... since I “own” both sides of this application (and can, if need be, add instrumentation or even test-jigs to “the back side”), and since the app is thought to be “working now,” my primary purpose in this case is somewhat specialized.   I want to establish how the app is supposed to be “working” now, so as to be able to automatically detect when it isn’t working any more, i.e. as the unintentional result of a recent change that introduced a bug.   In other words, regression testing.   So I will be mainly looking for differences from a known set of outputs given a known set of simulated-user inputs.   I want to be able to accomplish this automagically and at high speed.   Any report of discrepancies can trigger (manual, human) investigation, which will not need to involve automated assistance.