It depends on if by "immutable" you mean the only operation you are performing on the dataset are reads. Writes is a catch-all term usually used to describe ei

It depends on if by "immutable" you mean the only operation you are performing on the dataset are reads. Writes is a catch-all term usually used to describe either updates or inserts. If you are inserting new data and a single chunk is hot because a you are inserting a lot of data into it, then replicating won't help. You can imagine a scenario like a single device is going haywire and starts sending you a ton of data points. If you are only performing reads on your dataset, then replicating will only improve performance.