Have you never needed to merge datasets, that grew separately, in parallel? It is a royal pain when the rows have consecutive sequential primary IDs. Something

Have you never needed to merge datasets, that grew separately, in parallel? It is a royal pain when the rows have consecutive sequential primary IDs. Something has to make way for the other. IMO, this is one thing Mongo got right: keys, by default, are a timestamp + some random string. This way, they still sort nicely, but don't collide with their cousins in a different instance of the DB.