Currently each put operation overwrites all measurements for a given key. We will shortly be introducing PATCH style operations, allowing you to overwrite specific measurements.
Currently data is aggregated as a background process. Tunable consistency will allow you to specify if data should be aggregated and written in the same transaction as the raw data.
Currently we support "classic" time series data along with interval data. Our experience in the energy sector tells us that native support for cumulative data will be a great addition.
We believe there is scope for significant performance improvements, for both ingestion and querying.