Some tests depend on changes made to the environment by previous tests. This is Bad.
This bug is a prerequisite for running the tests in "make test" in parallel, since that will break the order in which they're run (since subsequent tests might be run by different workers).
In order to prevent this kind of dependency, we should run tests in a random order.
We already have a --randomize-order flag to selftest to do this. However, we need to fix the existing tests to cope with randomized order before we can turn on this flag everywhere.
Created attachment 10347 [details]
Subunit output when running with --random on sn-devel-104
The current list of failed tests are: