Currently, we maintain two versions of the retryable write pre-fetch pipeline, one is in c++ code for production and another one in javascript for testing. We should only need to maintain one copy and maybe we can modify the javascript test to query the oplog buffer to make sure the correct entries are in there.