We have a series header with normalization interval set to MINUTE. We also have two years of historical data and potentially billions of rows. If we run tests and realise that there is no sense normalizing this data to MINUTE level and HOUR or in some cases DAY will suffice and we then change the interval on the ts header to HOUR or DAY, …
does that mean we need to re-normalize the entire series with the new interval?
What would happen if it was left as is, and we queried a time range that spans data for which some data is normalized at a different interval to the rest?
Trying to avoid having to re-normalize a massive amount of data.