Writing out of order timestamps to the history store causes visibility problems that are difficult to resolve. There are very few cases in MongoDB where they can use timestamps out of order. We should be able to keep the out of order content in cache until the section of the update chain that is out of order becomes globally visible and can be discarded.
That's only going to be OK if the content isn't between the stable and oldest timestamps. i.e: checkpoint doesn't need to flush the content that has a range of out of order timestamps.