I've got a family of tests that have just developed a pathology. @Before creates a service on the local transport and a WebClient set to talk to it.
@After shuts it down. I can run any single test as many times as I like, all is well. If I use IntelliJ or maven-surefire to run the whole bag, a few will fail with symptoms indicating that the wrong request ended up at the service; one from a prior or different test case. The failure moves around from test to test from run to run. I've seen this with 3.1.1 and 3.1.2. We do have DIRECT_DISPATCH turned on. This whole thing was working in the middle of yesterday, and I seem to have, well, breathed on it. I certainly changed the timing by eliminating a lot of noisy logged backtraces.
