The test failure is in radv, in test_cs_uav_store(). I think that's pretty definitely unrelated. I'm not really sure I can reasonably debug it.
Given the results we're seeing (i.e., we seem to be reading back the clear value instead of the expected value from the dispatch), there's a perhaps conspicuous lack of uav_barrier() calls in test_cs_uav_store(). Is there anything in d3d12 that allows us to omit these? On the Vulkan side, UAV clears are essentially vkCmdDispatch() calls.
The d3d12 documentation is less than complete. It uses the language "all UAV accesses (read or writes)" when talking about UAV barriers, but then only mentions "draw or dispatch calls" specifically. It's not really clear whether clears count. Has this test ever failed on Windows? Should we just add a barrier there anyway?