Open
Description
In the velocity aggregated product, we cycled through all the vertical cells at a given timestamp before moving on to the next timestamp. In the velocity hourly, we add each "slice" of timeseries (currently up to 90 days long) at one cell, then the cycle through the cells.
Given that we're building a properly indexed ragged array, the output timeseries can be reconstructed the same way in each case (via the instrument_index and CELL_INDEX variables), but it might be helpful to be consistent.
This was pointed out in #99 (comment), and I don't think we ever addressed it.