30 minutely tides across a three year epoch and our ~10 ensemble models is a lot of computation, and currently takes up to 15 minutes of our processing time for a tile.
It would be great to do a sensitivity analysis of how much reducing our frequency from once every 30 minutes to once per hour would impact our results. If there was little impact, this would be an easy way to reduce our processing costs.