I forgot the exact scenario, maybe during a CK_FORK=no, but if I remember correctly there is a test before that, that need to trigger an ecore_shutdown, which lead to this test having no init when started. I will have to look at it more to remember.
The way the promise test is written doesn't follow that pattern at all. To maintain the count correct for ecore_init in other test the fixture does shutdown ecore during the fixture init and turn it back in after. The change above just follow the current pattern used in this file. If we want to change this, we will have to patch the entire file.