My concern with a memory dataset tag is that they are stored in the internal database.
If the size isn't too great then it probably isn't a major concern.
Technically Datasets and Databases aren't ordered either, you have to enforce that order. The same can be done with JSON.
With the size of LOAD you're currently looking at probably not an issue with any direction that you choose, however, if there is ever going to be the intent to scale this up, then you might want to plan for that.
I think I would agree that if you have a database at your disposal then that might be your best option. An added bonus is that you can also then have a historical record.