We need to know how to estimate the cost of the query when it's evaluated before we use those resources.

Sample limit might be not enough, as cost of fetching 1000 samples for single series is different than 1000 series for one sample.

• What is the unit/resource we can limit / configure limit?