With test.
test_xmlrpc has a timeout detection code which is simply broken (and it's actually documented): I just removed it, so if the server loops, the test will block. I think it's acceptable since other tests behave in the same way, and those days we have faulthandler that can be used to pinpoint such deadlocks/loops easily. Also, I've noticed that people are more inclined to fix tests that block than mere failing tests :-)