The test sample_rates_bulk_write_multi_namespaces.js is reliant on hitting a target throughput of commands executed to get a sample rate. Windows machines seem to be a bit slower and cause this test to fail inconsistently due to not hitting the threshold executions / second. Rather than further reduce the numbers which will impact how useful the test is on linux it makes more sense to disable the test on windows