Even after WT-8401, test_insert_over_delete_replace in test_inmem01 still fails occasionally on FLCS. I'm reasonably certain the analysis in WT-8401 was sound, but it seems that the threshold for skipping the test isn't as portable as I'd thought/hoped, and until someone (who has more access to the test environment than I do) wants to take the time to figure out a suitable threshold for the test environment and/or how to conditionalize it, the best approach for the time being seems to be to just always skip the test. Note that much of the scenario is covered by the separate test_insert_over_delete subtest, so the coverage loss is not high.