The D3DKMTCheckOcclusion() tests had some more occlusion failures after the commits from bug 5131:
* 2023-02-01 on w1064v1607 (32-bit) * 2023-02-01 on w1064v1607 (64-bit) * 2023-02-06 on w864 (64-bit) along with very many other occlusion failures
So this merge request marks them as flaky too. But then we did not get any new failures in the past 10 days. So now I'm not entirely sure this is needed after all: maybe the source of the occlusions got fixed in some other test?