Consistently high CPU usage on hot nodes

I was unsure if I can just adapt the index template and rollover. So there is no issue with a data stream having indices of type time-series and logsdb?

I just want to avoid creating an additional problem.