Reading further, the pricing appears to be calculated as  `price = write_per_sec_at_95_percentile * num_datapoints`. So, despite the storage of data is free and the new datapoints don't as many write, the pricing will still be the top write_per_sec from the other part. Plus the addition of the new datapoints, the cost would increase drastically.

So, it appears not to be a good idea to implement in this way as for higher cost.