On #19583 we reviewed test results on Jenkins and found cases where even less of the write cache was covered by the pattern. The lowest was 86%. We decided to allow those tests to pass to have fewer irrelevant test failures on Jenkins.